Database tool halves design iterations for lightweight, high-performance composites
Challenge
To streamline the development of Tailored Fibre Placement (TFP) components by creating a parameter database that enhances design understanding and manufacturing efficiency, validated through a complex TFP demonstrator.
Background
Composites are increasingly used for lightweighting across sectors, and Tailored Fibre Placement (TFP) technology offers a way to produce high-performance custom preforms with greater design flexibility than traditional methods.
TFP allows fibres to follow complex shapes by using a zigzag stitching technique to secure fibres to a backing fabric and minimise fibre damage, thus reducing waste and costs while enhancing automation. It ensures precise fibre orientation for optimal mechanical performance and supports various reinforcement materials, including carbon, glass and ceramic fibres. TFP is emerging as a technology that has the potential to achieve particularly light components for a given performance.
However, TFP applications have limitations; achieving dimensional accuracy is difficult due to variables like fibre stacking, compaction and stitching inconsistencies. Moreover, a limited understanding of how design parameters affect manufacturing makes developing TFP parts time-consuming and expensive, and there's a lack of clear data on potential mass reduction and performance gains.
This research, funded by the High Value Manufacturing (HVM) Catapult, seeks to address these challenges by developing a TFP parameter database to improve design understanding and enable faster, data-driven decisions, demonstrated through the design and manufacture of a complex TFP component.
Innovation
University of Sheffield AMRC experts conducted a study to identify and analyse the effects of key input and output variables for the TFP process, implementing a Design of Experiments (DoE) approach – a method to efficiently identify the most influential variables from a larger set by using a reduced number of experimental runs.
This approach, applied to eight input variables, resulted in a carefully planned set of 54 experiments (a ‘54-run plan’). For each of these 54 experiments, samples were designed in EPCwin and manufactured as symmetrical two-layer panels with specific fibre orientations. All data was recorded and analysed using Minitab Statistical Software, while ANOVA and Pareto charts determined statistical significance, with R2 values assessing model fit.
For trial efficiency, a separate study assessed the impact of layer count and lay-up type by manufacturing 36 preforms with varying configurations and measuring their thickness using a Hans-Schmidt gauge.
The combined results from the DoE and thickness studies were used to develop a predictive model for TFP processes, with key analyses to identify significant factors, assess model accuracy and ensure data reliability.
Data from both studies was used to create a TFP component parameter database for future projects, allowing users to input design variables and predict the output to enable faster iterations without physical prototyping.
The database was validated by manufacturing a racing car wheel, a structural component, with optimisations in part thickness and fibre orientation. The iterative design process was faster than previous methods, and the final component used a stabilised construction of dry TFP and non-crimp fabric (NCF) preforms.
Result
DoE study:
The study identified key TFP parameters influencing the final product: thickness was mainly affected by filament count, backing fabric, stitch length and carbon offset; and area deviation by pull compensation limit, stitch length and edge stitch length.
The DoE analysis showed high R² values, indicating a strong predictive model, and residual analysis confirmed generally normal data distribution with some outliers, aiding design and manufacturing optimisation.
Thickness study:
This study examined how carbon fibre preform thickness varies with layer count. As expected, more layers increased thickness – 12k fibre preforms were thicker than 6k. However, carbon offset variations caused unexpected thickness results in 12k fibre, while ply lay-up had minimal impact on thickness across preforms with the same layer count.
TFP parameter database:
The project successfully tested a new database tool that refined the TFP design process – reducing design iterations from six to three through precise adjustments of parameters like carbon offset and stitch length.
However, it couldn't entirely eliminate trial-and-error, particularly for complex shapes where thickness in curved geometries tends to be underestimated.
Furthermore, to address quality assessment, a new preform quality score was developed, offering a standardised and flexible ranking method, though its accuracy faced limitations with geometry and measurement.
Impact
The study's findings lay the groundwork for integrating predictive modeling and quality control within a Product Lifecycle Management (PLM) environment. Connecting TFP process knowledge with platforms like 3DS 3DEXPERIENCE can facilitate a data-driven approach to developing lightweight, high-performance composite components.
To maximise the impact of these advancements, a development plan involving the Research Software Engineering team at the University of Sheffield is recommended, potentially leading to licensing opportunities in industries such as automotive, aerospace and sporting goods through workshops and webinars.
To further improve the tool, the AMRC team also suggested several areas for future research:
- Investigate a broader range of geometries
- Explore material variability, including different fibre types, tow tensions, and resin systems
- Implement an iterative feedback loop between simulations and physical trials
- Integrate the tool with computer-aided design and manufacturing systems
- Enhance the user interface
- Validate the tool across various manufacturing environments
- Apply the tool to multifunctional preforms.