Machine Learning based prediction of the effect of defects in the Automated Fiber Placement - MEMÒRIA

Aug 5, 2026

Upload by Jiven

The thesis develops a Machine Learning model to predict how lay-up defects affect the effective stiffness of composite laminates produced with Automated Fiber Placement. Components made by this technology are still inspected manually, and defect effects are typically estimated using simulation tools. The proposed model uses geometric defect characteristics as inputs and targets high-precision predictions of laminate effective stiffness. Training requires extensive defect configurations and motivates a parametrized finite element model at coupon level, validated by the reported multilayer perceptron results.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages96

Reading Time4 h 2 min

AccessFree to download

Acknowledgement

Resum

Resumen

Abstract

Contents list

Treball de Fi de Màster Universitari en Enginyeria Industrial Machine Learning based prediction of the effect of defects in the Automated Fiber Placement MEMÒRIA Màster lay-up Autor: Marcal Iborra Escarré Director: Javier Anyeto Gubert Ponent: - Convocatòria: Gener 2023 Escola Tècnica Superior d’Enginyeria Industrial de Barcelona Acknowledgement I would like to extend my sincere gratitude to the Technical University of Munich and the Department of Composite Materials for their unwavering support in the completion of my engineering project. Their invaluable guidance, resources, and expertise helped me to overcome numerous challenges and achieve a successful outcome. I am truly grateful for the opportunities they provided me with and I am confident that the knowledge and skills I have acquired will serve me well in my future endeavors. I am honored to have been associated with such a prestigious institution, thank you for your ongoing support. Iam especially grateful to Fabian Diemar, who supervised this project and pushed me to work with scientific rigor and presenting the results in a clear and precise way. He gave me all the advises and guidance that I needed, which helped me during the project to obtain the final project results I would like also to thank my family and my friends for all the support received during the project development. Specially to Roger Aylagas, Christian Fernandez and Enric Gil for their helpful suggestions in the many hours of discussion in the field of artificial intelligence and machine learning, which have helped me to reach a deep understanding on this field.. Resum L'ús del Automated Fiber Placement està estenent-se en la indústria aeroespacial. Lanecessitat de fabricar components estructurals compostos grans i complexes, fa que l'úsd'aquesta tecnologia sigui molt més eficient que la fabricació convencional amb col·locació manual. No obstant això, aquests components encara s'estan inspeccionant manualment i es calcula l'efecte dels defectes trobats amb software de simulació . L'abast d'aquesta tesi és crear un model de Machine Lerning que sigui capaç de calcularl'efecte en la rigidesa efectiva per diferents configuracions de defectes. Aquest model d'aprenentatge automàtic hauria de rebre les característiques geomètriques dels defectes en el laminat i de ser capaç de predir, amb un alt nivell de precisió, la rigidesa efectivadel laminat. Entrenar aquest model amb una gran quantitat de configuracions de defectes diferents genera la necessitat de crear un model FE parametritzat d'una laminació composta en el nivell de cupó . Els resultats mostren que una arquitectura de Multilayer Perceptron amb dues hidden layers. La primera amb 281 nodes i la segona amb 76 nodes, és capaç de predir la rigidesa efectiva d'un laminat defectuós amb una precisió de 0,1 GPa. Resumen El uso del Automated Fiber Placement se está expandiendo en la industria aeroespacial. La necesidad de fabricar grandes y complejos componentes estructurales de materiales compuestos, hace que el uso de esta tecnología sea mucho más eficiente que la fabricación manual convencional. Sin embargo, estos componentes siguen siendo inspeccionados manualmente y se calcula el efecto de los defectos encontrados con un software desimulación. El objetivo de esta tesis es crear un modelo de Machine Learning que sea capaz decalcular el efecto sobre la rigidez efectiva para diferentes configuraciones de defectos. A este modelo de aprendizaje automático se le deben proporcionar las características geométricas del defecto en el laminado y tiene que ser capaz de predecir, con un alto nivel de precisión, la rigidez efectiva del laminado. El entrenamiento de este modelo se debede realizar con una gran cantidad de configuraciones de defectos diferentes. Este hecho genera la necesidad de crear un modelo de elementos finitos parametrizado de un laminado a nivel de cupón. Los resultados muestran que una arquitectura Multi

FAQs

What problem does the thesis address?

It addresses the need to estimate how defects in Automated Fiber Placement affect the effective stiffness of composite laminates, given that inspection is still manual and defect effects are currently computed via simulation software.

How does the proposed Machine Learning model make predictions?

It takes geometric characteristics of defects in the laminate as inputs and predicts the effective stiffness of the laminate with high accuracy for different defect configurations.

Why is a parametrized FE model needed?

Training the Machine Learning model with a large variety of defect configurations requires generating data, which motivates building a parametrized finite element model for composite laminates at the coupon level.

Aug 5, 2026

Upload by Jiven