Date of Award

Fall 2023

Document Type

Open Access Thesis


Mechanical Engineering

First Advisor

Ramy Harik


Automated Fiber Placement is an advanced manufacturing technique for industrial-scale composite structures. Advanced robotics coupled with composite manufacturing results in faster and more consistent results than previously obtained through hand layup. The complexity and interconnectedness of the automated fiber placement process provides a difficult challenge for traditional modeling techniques. Modeling within automated fiber placement currently utilizes physics-based modeling to inform the translation of a design to a manufacturing plan. The intricacy of the automated fiber placement process dictates that attempts at modeling or optimizing these processes are often limited in their scope. Physics-based modeling for manufacturing typically involves numerous interacting systems and a variety of adjustable parameters that must be accounted for in the model. To account for the many degrees of freedom, arduous simulations are required. Even under the best computing environments, there is no guarantee that the frequent assumptions needed to create physics-based models have not abstracted away detail to a detrimental degree. Data-driven models, however, can provide elucidations where traditional analytic solutions have under-performed. This thesis will produce a comprehensive DDM base for AFP process parameters that will be utilized in a hybrid physics-data modeling framework for implementation into existing post-processing software. The primary utilization of DDM will be in the production of a heat transfer model for the prediction of the actual applied temperature and inversely the prediction of the optimal processing temperature. The application of heat is a key factor in ensuring proper adhesion between the substrate and incoming tows, and due to its black box nature and high variability, this situation lends itself perfectly to DDM. Data will be collected through experimentation for development of the models and a validation experiment will be performed to access the validity of the DDM models and the full hybrid-physics model. To demonstrate the benefits of such an integration, an in-depth validation will be performed on a complex curvature tool.


© 2024, Matthew John Godbold