Method for an Automated Optimization of Fiber Patch Placement Layup Designs

Benjamin Fischer, Bernhard Horn, Christian Bartelt, Yannick BlößlInternational Journal of Composite Materials 2015 · DOI

TL;DR

This paper introduces an automated method using an ant colony algorithm to optimize the placement of fiber patches in composite manufacturing. This approach addresses the limitations of traditional methods in creating complex, multi-curved structures, enabling better load path optimization and potentially reducing manufacturing costs and material waste.

What Problem Does This Paper Address?

The manufacturing of composite materials, especially those with complex geometries, is often expensive due to high material costs and significant manual labor. Existing automated techniques like filament winding, braiding, and fiber placement struggle with intricate shapes and small radii of curvature, leading to limitations in design flexibility and the ability to optimize material usage for maximum strength. Fiber patch placement (FPP) offers a solution by building structures from small, dry fiber patches, allowing for more adaptable fiber orientation. However, the sequential placement of these patches creates discontinuities that affect mechanical properties. Optimizing the distribution and arrangement of these discontinuities is crucial for maximizing the composite's strength, but the sheer number of possible patch configurations makes manual or brute-force optimization computationally intractable. This paper tackles the challenge of efficiently finding the optimal patch arrangement to exploit the full potential of FPP for complex structures.

Methods

The core of this paper is the development of an automated optimization method based on the ant colony algorithm. This heuristic approach is chosen for its ability to find good solutions for complex, NP-hard problems without requiring extensive preliminary data.

The method involves:

  • Fitness Function: A crucial component is the "quality parameter" (QP) that quantifies the laminate's quality, correlating with its expected tensile strength. This function is derived from geometrical parameters of patch overlaps and validated against experimental data.
  • Ant Colony Algorithm Adaptation: The natural ant colony behavior of finding the shortest path using pheromones is adapted. In this context:
  • Ants: Represent virtual agents exploring different patch allocation possibilities.
  • Paths: Correspond to sequences of patch movements along a "patch line" (a series of patches in a sub-layer).
  • Pheromones: Are dynamically adjusted based on the quality of the generated patch allocations. Higher quality allocations deposit more "pheromone," guiding subsequent ants towards better solutions.
  • Evaporation: Pheromones gradually decrease over time, preventing premature convergence to suboptimal solutions.
  • Pheromone Function: A specific pheromone function is developed to scale the fitness function's output into a usable range (1-100) for the ant colony algorithm, addressing issues with very small or very large fitness values.
  • Algorithm Sequence: The algorithm is structured into several steps: initialization, ant-driven patch allocation generation, quality calculation (fitness function), best solution storage, pheromone update, and loop termination.

Datasets

Not discussed in the paper. The paper focuses on a method and its validation using experimental data for the fitness function, rather than training on specific datasets.

Key Insights and Ideas

  • Distributing Discontinuities is Key: The paper highlights that the strength of FPP laminates is heavily influenced by how fiber discontinuities (gaps between patches) are distributed. Shifting patches to avoid aligning discontinuities across layers is crucial for maximizing strength, as opposed to simply placing them directly on top of each other.
  • Ant Colony Algorithm for Complex Optimization: The ant colony algorithm proves to be an effective heuristic for tackling the combinatorial explosion of patch placement possibilities, offering a computationally feasible approach where analytical methods would fail.
  • Fitness Function Validation: The reliance on experimentally validated geometrical parameters for the fitness function lends credibility to the optimization process, ensuring that the algorithm aims for physically relevant improvements in strength.
  • Performance Optimization is Crucial: For practical application, the authors emphasize the need for performance optimizations like caching calculations to make the algorithm viable within reasonable timeframes, especially for complex components.

Key Results

  • The proposed method, using an ant colony algorithm, is demonstrated to be significantly more effective than a random search algorithm for optimizing patch placement.
  • In an example with a complex, multi-curved medical orthosis sole, the ant colony algorithm produced acceptable solutions in 5 out of 10 optimizations, compared to only 1 out of 10 for a random algorithm.
  • The algorithm's performance improves with an increased number of loops, suggesting that longer optimization runs lead to better quality solutions.
  • The developed algorithm is shown to be much faster than analytical methods for finding optimal patch allocations, with a practical example illustrating the immense time savings.

Limitations and Future Work

  • Computational Cost: While faster than analytical methods, the algorithm can still be time-consuming, especially for very complex components or a high number of loops. The authors suggest further performance optimization through caching and parallelization.
  • Empirical Pheromone Function: The pheromone function's scaling factors are determined empirically. Further research could explore more robust or adaptive methods for this scaling.
  • Generalizability: While validated on a specific example (medical orthosis sole), further testing on a wider range of complex geometries would be beneficial to fully assess the algorithm's generalizability.
  • Future Work: The authors suggest adapting the algorithm to run in parallel to further increase performance and developing the entire manufacturing process chain for better efficiency.

Summary updated

Abstract

The manufacturing of composite structures is expensive due to high material cost and the amount of manual labor needed. Current manufacturing technologies, e.g. filament winding, braiding or fiber placement technologies are offering a possibility of automated manufacturing to lower these costs. Nevertheless, these technologies are limited when it comes to structures with complex fiber architecture and small geometries. A new approach to overcome these limitations is the fiber patch placement technology. The preform of a composite structure is built up with small and dry fiber patches in a sequential order. Fiber discontinuities are necessarily occurring between two patches and are influencing the mechanical properties. To optimize strength of the composite those discontinuities have to be distributed within the preform. With increased number of patches, the number of combinations of patch positions is increasing significantly. This paper presents a method based on an ant colony algorithm for an efficient calculation of an optimized distribution, which enables to use the potential of patched laminates.
BibTeX
@article{fischer2015method,
  author = {Benjamin Fischer and Bernhard Horn and Christian Bartelt and Yannick Blößl},
  title = {{Method for an Automated Optimization of Fiber Patch Placement Layup Designs}},
  journal = {International Journal of Composite Materials},
  year = {2015},
  doi = {10.5923/j.cmaterials.20150502.03},
}
Figure