Study on the effect of reworking Gap/Overlap defects on the quality of AFP-produced laminates
Topic(s) : Manufacturing
Co-authors​ :
Stig MCARTHUR (UNITED KINGDOM), Jorn MEHNEN , Iain BOMPHRAYAbstract :
This paper delves into the assessment of reworking techniques in the realm of composite materials, aiming to transcend conventional cosmetic pass-fail criteria commonly employed by practitioners. Acknowledging the limited scope of such criteria in capturing the enduring impact of defects post-rework, our study focuses on laminates, drawing comparisons among pristine samples, samples with embedded defects, and those subjected to the reworking process.
Our experimental approach, utilizing a mock-Automated Fiber Placement (AFP) setup, establishes a robust research environment for simulating real-world conditions. The successful generation of laminates using this method demonstrates its potential as a democratised research environment free of the high capital costs traditionally associated with AFP research and development.
Three distinct sample classes were manufactured: 1) Pristine samples devoid of defects; 2) Defective samples featuring artificially induced gap/overlap defects in the central ply; and 3) Reworked samples, which featured an identical defect configuration in the central ply as the Defective samples. These defects were then reworked following the guidelines in literature and practice, resulting in a laminate with as close to pristine as possible.
A robotically controlled Ultrasonic Non-Destructive Testing (NDT) method was employed, with a deep learning approach for data interpretation, in order to study the laminates post-cure. Our study utilized a linear phased array ultrasonic roller probe to capture B-scans across sample lengths. These B-scans were subsequently concatenated into C-scans, providing a continuous cross-sectional view of the laminate's internal structure.
The information extracted from these scans facilitated a detailed comparison of internal structures among the different sample classes. Despite the successful completion of the rework process, a discernible difference between pristine and reworked samples, termed the "rework signature," was identified. This observation underscores the critical need for optimizing reworking activities from a quality perspective.
Furthermore, our study advocates for additional mechanical tests to thoroughly examine the impact of these rework signatures on the overall performance of composite parts. By integrating advanced techniques such as robotically controlled Ultrasonic-NDT and deep learning-based data interpretation, this research contributes not only to the understanding of rework effectiveness but also emphasizes the importance of a comprehensive approach to quality assessment in composite material manufacturing. The findings pave the way for future optimization strategies, emphasizing the necessity of considering both internal and surface characteristics in evaluating the integrity of composite materials.
Our experimental approach, utilizing a mock-Automated Fiber Placement (AFP) setup, establishes a robust research environment for simulating real-world conditions. The successful generation of laminates using this method demonstrates its potential as a democratised research environment free of the high capital costs traditionally associated with AFP research and development.
Three distinct sample classes were manufactured: 1) Pristine samples devoid of defects; 2) Defective samples featuring artificially induced gap/overlap defects in the central ply; and 3) Reworked samples, which featured an identical defect configuration in the central ply as the Defective samples. These defects were then reworked following the guidelines in literature and practice, resulting in a laminate with as close to pristine as possible.
A robotically controlled Ultrasonic Non-Destructive Testing (NDT) method was employed, with a deep learning approach for data interpretation, in order to study the laminates post-cure. Our study utilized a linear phased array ultrasonic roller probe to capture B-scans across sample lengths. These B-scans were subsequently concatenated into C-scans, providing a continuous cross-sectional view of the laminate's internal structure.
The information extracted from these scans facilitated a detailed comparison of internal structures among the different sample classes. Despite the successful completion of the rework process, a discernible difference between pristine and reworked samples, termed the "rework signature," was identified. This observation underscores the critical need for optimizing reworking activities from a quality perspective.
Furthermore, our study advocates for additional mechanical tests to thoroughly examine the impact of these rework signatures on the overall performance of composite parts. By integrating advanced techniques such as robotically controlled Ultrasonic-NDT and deep learning-based data interpretation, this research contributes not only to the understanding of rework effectiveness but also emphasizes the importance of a comprehensive approach to quality assessment in composite material manufacturing. The findings pave the way for future optimization strategies, emphasizing the necessity of considering both internal and surface characteristics in evaluating the integrity of composite materials.