Abstract

Automated Fiber Placement (AFP) has become a critical manufacturing technology for large-scale composite aerospace structures. However, despite advances in machine automation, AFP process planning remains largely dependent on expert-driven trial-and-error, particularly for complex geometries. This limitation results in inconsistent quality, high programming costs, and extended development cycles. This study proposes a defect-aware Computer Aided Process Planning (CAPP) framework that reformulates AFP programming as a multi-objective optimization problem. The framework integrates ply-level defect prediction, Analytical Hierarchy Process (AHP)-based weighting, scenario generation, and dynamic layup strategy selection. Geometry-driven defects—including gaps, overlaps, angle deviation, and steering violations—are quantified and evaluated across multiple candidate scenarios. Predicted defects are validated against inspection data, revealing strong correlation trends while highlighting limitations of geometry-only modeling. Furthermore, a through-thickness defect stacking model is introduced to evaluate cumulative defect risk across laminates. Results demonstrate that AFP process planning can be significantly improved by shifting from static, experience-based decision-making to adaptive, defect-prioritized optimization. The proposed framework establishes a pathway toward closed-loop, geometry-aware AFP manufacturing systems.

Keywords

Automated Fiber Placement; Computer Aided Process Planning (CAPP); Defect Prediction; Dynamic Layup Strategy; Analytical Hierarchy Process (AHP); Composite Manufacturing; Gap and Overlap; Fiber Steering; Process Optimization

1. Introduction

Automated Fiber Placement (AFP) has emerged as a dominant manufacturing method for high-performance composite structures used in aerospace systems, including cryogenic tanks, fuselage sections, and load-bearing panels. Its ability to deposit narrow tows along controlled trajectories enables efficient fabrication of complex geometries with tailored fiber orientations. Despite these advantages, AFP process planning remains one of the least automated stages in the manufacturing chain.

In current industrial practice, process planning relies heavily on the experience of skilled engineers who manually define layup strategies, starting points, and path propagation rules. This approach is inherently iterative and often leads to inconsistent results, particularly when applied to doubly-curved or geometrically complex surfaces. Small variations in path definition can produce significant differences in defect formation, directly impacting both manufacturing efficiency and structural performance.

A critical observation is that many AFP defects are not stochastic in nature but are deterministic outcomes of geometric constraints. Gaps, overlaps, and excessive steering arise from the interaction between fiber paths and surface curvature. However, existing workflows treat these defects as inspection outcomes rather than as controllable variables during process planning.

This study addresses this gap by introducing a defect-aware AFP process planning framework. The central premise is that AFP programming should be treated as a multi-objective optimization problem, where candidate layup scenarios are generated, evaluated based on predicted defect behavior, and selected according to application-specific priorities. By embedding defect awareness into the planning stage, the framework aims to reduce reliance on manual iteration and improve overall manufacturing consistency.

2. Literature Review

AFP research has historically focused on three primary areas: path generation algorithms, defect characterization, and process simulation.

Path generation methods, including geodesic, parallel, and constant-angle strategies, have been widely studied to minimize fiber distortion and steering. While effective for simple geometries, these approaches often fail to address the competing constraints present in complex curvature fields. No single layup strategy has been shown to consistently minimize all defect types.

Defect characterization studies have identified a broad spectrum of AFP defects, including gaps, overlaps, wrinkles, puckers, bridging, fiber angle deviation, and tow misplacement. Traditional planning approaches prioritize gap and overlap minimization, often neglecting other defect modes that may be equally critical in certain applications.

Simulation tools such as VERICUT Composite Programming have enabled virtual prediction of certain defects during planning. However, these tools are typically used for verification rather than optimization. The integration of defect prediction into decision-making remains limited.

Additionally, most existing studies evaluate defects at the ply level, without considering their interaction through laminate thickness. In practice, defect stacking across multiple plies can significantly amplify structural risks.

Overall, current literature lacks a unified framework that integrates defect prediction, prioritization, validation, and adaptive path generation within a closed-loop planning system.

3. Methodology

3.1 Framework Overview

The proposed framework is built around a Computer Aided Process Planning (CAPP) architecture that integrates geometry processing, scenario generation, defect evaluation, and decision-making. The workflow consists of:

1. Generation of candidate layup scenarios
2. Extraction of defect metrics for each scenario
3. Weighted scoring using Analytical Hierarchy Process (AHP)
4. Selection of optimal scenarios based on defined priorities

3.2 Defect Classification

AFP defects are categorized into two primary groups:

• Geometry-driven defects, including gaps, overlaps, angle deviation, and steering violations
• Manufacturing-driven defects, including wrinkles, puckers, bridging, and tow instability

The framework focuses on geometry-driven defects, which can be predicted during planning, while acknowledging the limitations in capturing manufacturing-induced variability.

3.3 Ply-Level Defect Quantification

Each candidate layup scenario is evaluated using quantitative defect metrics:

• Gap area distribution
• Overlap area distribution
• Excess fiber angle deviation
• Minimum steering radius violations

These metrics are extracted from virtual simulations and normalized to enable comparison across scenarios.

3.4 AHP-Based Defect Weighting

The Analytical Hierarchy Process (AHP) is used to assign relative importance to different defect types. Through pairwise comparisons, process planners can encode application-specific priorities, such as emphasizing structural integrity over manufacturing efficiency or vice versa.

The weighted scoring function is defined as:

Total Score = Σ (Weight_i × Defect Metric_i)

This allows the framework to identify optimal solutions based on context rather than assuming uniform defect importance.

3.5 Scenario Generation

Multiple layup scenarios are generated by varying key planning parameters:

• Layup strategy (e.g., parallel, geodesic, adaptive)
• Starting point selection
• Path propagation direction

This creates a solution space in which trade-offs between competing defect mechanisms can be explored.

3.6 Experimental Validation

Predicted defect distributions are validated against inspection data obtained from laser profilometry systems. The comparison focuses on:

• Spatial alignment of predicted and observed defects
• Total defect area correlation
• Identification of discrepancies between virtual and physical results

3.7 Through-Thickness Defect Stacking

A discretized stacking model is introduced to evaluate defect accumulation across laminate thickness. This model identifies regions where repeated defect occurrence may lead to structural degradation, such as resin-rich zones or local stiffness reduction.

3.8 Dynamic Layup Strategy

A dynamic layup strategy is developed by combining multiple propagation approaches and adjusting path generation based on defect weighting. Unlike static strategies, this method adapts to local geometry and evolving defect conditions.

4. Results

The results demonstrate that geometry-driven defect prediction successfully captures the primary trends of gap, overlap, and angle deviation distribution across complex surfaces. Comparison with inspection data shows strong spatial correlation, with discrepancies primarily attributed to manufacturing effects such as tow deformation and compaction variability.

Quantitative analysis indicates that overlap area prediction achieves reasonable accuracy, with deviations reflecting the inherent limitations of geometry-only models.

Scenario-based evaluation reveals that different layup strategies produce distinct defect patterns, and no single strategy consistently minimizes all defect types. The AHP-based scoring framework enables effective selection of scenarios aligned with specific engineering priorities.

Through-thickness analysis shows that defect stacking significantly increases local defect severity. Variations in starting point selection and path propagation can reduce defect co-location and improve laminate quality.

Dynamic layup strategies demonstrate improved performance compared to traditional fixed strategies, particularly in reducing excessive steering and angle deviation while maintaining acceptable gap and overlap levels.

5. Discussion

The findings confirm that AFP process planning is inherently a multi-objective optimization problem. Defect mechanisms are interdependent, and efforts to minimize one type often lead to increases in another. As a result, optimal solutions must be defined relative to application-specific requirements rather than universal criteria.

The integration of AHP weighting provides a structured method for incorporating engineering judgment into the planning process. However, the effectiveness of this approach depends on the accuracy of the assigned priorities.

The validation results highlight the boundary between geometric prediction and physical reality. While geometry-driven defects can be predicted with reasonable accuracy, manufacturing-driven defects require more advanced modeling of material behavior, compaction forces, and thermal effects.

The introduction of through-thickness defect stacking represents a critical advancement, shifting the focus from individual ply quality to cumulative laminate behavior. This approach aligns more closely with structural performance requirements in aerospace applications.

Overall, the proposed framework represents a transition from static, experience-based planning to adaptive, data-driven AFP manufacturing.

6. Conclusion

This study presents a defect-aware AFP process planning framework that integrates defect prediction, weighted evaluation, experimental validation, and dynamic layup strategy generation within a unified CAPP system.

The key contributions include:

• Reformulation of AFP process planning as a multi-objective optimization problem
• Introduction of AHP-based defect prioritization
• Validation of defect prediction against inspection data
• Development of through-thickness defect stacking analysis
• Implementation of dynamic, adaptive layup strategies

The results demonstrate that defect-aware planning significantly improves AFP process efficiency and quality. Future work should focus on integrating manufacturing physics into defect prediction models and establishing direct links between defect metrics and structural performance.

References

1. Halbritter, J. A. (2023). Leveraging Automated Fiber Placement Computer Aided Process Planning Framework for Defect Validation and Dynamic Layup Strategies. University of South Carolina.
2. Croft, K., et al. “Automated Fiber Placement for Aerospace Structures.” Composites Part A.
3. Shirinzadeh, B., et al. “Path Planning in Automated Fiber Placement.” Journal of Manufacturing Systems.
4. CGTech. VERICUT Composite Programming (VCP) Documentation.
5. Advanced Composite Structures Inspection System (ACSIS) Technical Reports.