An article by TU Delft on their recent Deliverable 5.3 Topology optimization open-source code.
The use of fiber-reinforced polymer composites (FRPCs) has transformed industries such as aerospace, automotive, and renewable energy due to their excellent strength-to-weight ratio. Automated processes, such as Automated Fiber Placement (AFP), are gaining interest to harness the full potential of FRPCs by placing the materials only where they are needed and in the main load directions based on the structural design.
What is Automated Fiber Placement?
Automated Fiber Placement (AFP) is a manufacturing technology that uses a robotic head to automatically place fiber tows (reinforcing fibers) into a mold. This technique provides precise control over fiber direction and location, improving the quality and performance of the final laminate. Dry Fiber Placement (DFP) is a subclass of these processes where dry fibers are placed on the mold and then impregnated by the matrix polymer in a subsequent step in a closed mold.
Key advantages of fiber placement methods over other manufacturing methods include:
- Compatibility with topological optimization techniques, enabling the design of more efficient and lightweight structures with greater strength and stiffness.
- Reduced manufacturing time because the process is automated and significantly faster than traditional methods.
The positioning of the fibers in this process is critical, as placement inaccuracies can result in gaps or overlaps. These irregularities directly affect the permeability of the preform and can lead to the formation of voids (trapped air) during resin injection. This, in combination with the irregularities in the fiber directions, result in significant reduction in performance of structures.
The challenge arising from permeability variation due to dry fiber placement
In structural applications of FRPCs, plies are oriented at different angles and the variability in a single roving’s positioning and its width lead to non-uniform permeability. A local jump in the permeability will have a short range effect that will result in faster impregnation of the low fiber volume fraction areas (e.g. gaps between rovings) or slower impregnation of the high fiber volume fraction areas (e.g. overlaps), which introduces the risk of void formation that is not foreseen in the process design of the mold filling.
The CAELESTIS Methodology
In the CAELESTIS project, we have developed a methodology to predict and correct laminate permeability during the manufacturing process. Our solution is based on:
- Real-time monitoring: Using sensors and advanced inspection techniques implemented in the project, we can measure the positioning of the fiber tow during placement. Herein, we assume that technological advances within CAELESTIS, and other projects, will allow identification of gaps and overlaps when a layer is deposited.
- Predictive modeling: We use computational models developed in the project to predict what will be the permeability distribution due to variabilities introduced by the last deposited layer.
- Adjustment of subsequent layers: Based on the collected data, we propose a methodology to adjust the fiber positioning in the subsequent layers to homogenize the permeability of the laminate in order to mitigate the risk of void formation.
The proposed methodology is based on a machine learning generative model called Variational Autoencoder (VAE) and using a concept called variational alignment to bridge the permeability distribution and roving positions.
Advantages of the proposed methodology
As part of the project, a virtual manufacturing process was performed on a small computational domain. A large dataset (54000 samples) was generated using numerical models already developed within the CAELESTIS project. The hyperparameters of the VAE architecture were optimized during its training process.
After the VAE models were trained, the capabilities of the methodology were demonstrated by analyzing 10000 cross-ply laminates of 10 layers. From the proposed method, we can determine the position of the fiber tows of the subsequent two layers to homogenize the current permeability. We defined the methodology for the next two layers since it is the base unit of cross-ply laminates. From this analysis, we found that this methodology significantly improved the permeability uniformity of the laminate, resulting in several key improvements:
- Reduced void presence, minimizing internal defects and improving structural
- More uniform mechanical properties, eliminating weak zones within the structure.
- Increased material reliability and durability, essential for critical applications such as aerospace.
Automated Fiber Placement (AFP) is a key technology for manufacturing fiber-reinforced laminates, offering advantages in precision, efficiency, and structural optimization. However, challenges related to permeability can affect the quality of the final laminate. Thanks to the work carried out in the CAELESTIS project, we have implemented an innovative methodology that enables permeability homogenization and consequently improves the mechanical properties of the manufactured laminates. This research represents a step forward in the manufacturing of advanced composites, with applications in industries that require high technological standards.
Future research should focus on scaling the methodology to larger domains beyond those analyzed in this project.
