Spring-in and warpage of angled composite laminates

https://doi.org/10.1016/S0266-3538(02)00105-7Get rights and content

Abstract

The residual stresses that develop in fibre-reinforced laminates during autoclave processing while the laminate is confined to the process tool often lead to dimensional changes such as spring-in of angles and warpage of flat sections. A large number of experiments were performed to examine the effect of design and process parameters on spring-in and warpage of composite laminates with a symmetrical lay-up. The parameters studied include part angle, thickness, lay-up, flange length, tool material, tool surface, and cure cycle. This paper shows that both design and process parameters can have a significant effect on spring-in and warpage and that there are interactive effects between them. The paper also shows that spring-in of angled laminates is sensitive to the measurement technique as the measured spring-in often is compounded by warpage of flat laminate sections.

Introduction

During autoclave processing of fibre-reinforced polymer composite structures there is an inevitable build-up of residual stresses, mainly due to differential thermal expansion between fibres and matrix, which often causes distortions of the fully cured structure. For the simple L and C geometries studied here, these distortions can be classified as warpage, or curving of initially flat sections, and spring-in, or a reduction of enclosed angles of angled sections. Spring-in will routinely occur in angled autoclave processed parts whereas warpage may or may not be an issue, depending on the design of the structure and the process conditions. Spring-in causes problems in assembly because of poor fit-up to mating structures and is typically compensated for when designing the process tooling using an experience based compensation factor. The absolute magnitude of the spring-in, however, is difficult to predict and is often variable in production. Fig. 1 shows a schematic of the effect of spring-in of a spar flange on the fit-up of fastener locations between a spar and a wing skin. Any warpage (curvature) of the spar flange or wing-skin would exacerbate the fit-up problem. To quantify process-induced deformation for simple geometries such as the spar shown in Fig. 1, the concept of a spring-in angle is commonly used. However, as will be demonstrated in this paper, spring-in angles are often not sufficient to quantify process-induced deformations as the deformation is often compounded by warpage of flat sections.

Spring-in is not only affected by the differential thermal expansion between the fibres and the matrix, but also by the evolution of mechanical properties of the laminate during processing. While fibre properties remain essentially constant, resin properties change dramatically throughout the cure cycle as the resin polymerizes. The development of residual stresses is dependent on the processing history because of the large changes in mechanical and physical properties during cure. Thus, spring-in, which is the result of residual stresses developing while the composite part is confined to the process tool, depends on the timing of events during processing.

Many parameters are known or suspected to affect spring-in [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12] and it is convenient to classify them as intrinsic or extrinsic parameters. Intrinsic parameters are here defined as parameters related to part geometry and material properties, whereas extrinsic parameters are defined as parameters related to tooling and processing.

A simple equation [Eq. (1)] has been proposed for predicting spring-in of angled laminates based on material anisotropy [1], [2]. This equation accounts for the temperature difference between cure and ambient conditions, the anisotropy of thermal expansion and cure shrinkage, and the part angle Fig. 1.Δθ=ΔθCTE+ΔθCSαl−αtΔT1+αtΔTφl−φt1+φtwhere: Δθ=spring-in angle; ΔθCTE=thermal component of the spring-in angle; ΔθCS =cure shrinkage component of the spring-in angle; θ=part angle; αl =longitudinal coefficient of thermal expansion; αt=through thickness coefficient of thermal expansion; ΔT=difference between cure temperature and ambient temperature; φl=longitudinal cure shrinkage; φt=through-thickness cure shrinkage.

The first term of Eq. (1), ΔθCTE, is the thermal expansion anisotropy component, which is the result of residual stresses that develop during cool-down, when the laminate is fully cured. The second term, ΔθCS, is the result of resin cure shrinkage and is associated with stresses that build up earlier in the cure cycle, as the material cures. Published results indicate that Eq. (1) in some cases provides a reasonable estimate of spring-in [3], [4]. However, Eq. (1) does not account for extrinsic parameters such as tooling effects.

There is disagreement of the effect of corner radius on spring-in in the literature. One study indicates that spring-in is smaller for smaller radii [5]. Other studies have concluded that corner radius has little effect on the overall spring-in [6], [7]. The effect of part thickness on spring-in is also unclear. While some studies have found that thinner parts have greater spring-in than thicker ones [5], [8], other studies [6], [7] revealed little difference. In contrast to these findings, one study found that doubling the part thickness increased the spring-in by over 20% [9]. The differences between these findings suggest that the effect of thickness on spring-in may be sensitive to other interacting parameters.

It has been reported that ply orientation has little effect on spring-in when comparing cross-ply, angle-ply and quasi-isotropic laminates [6], [9]. It is generally agreed that spring-in is virtually zero for 90° lay-ups [6], [9], whereas there is disagreement regarding the magnitude of spring-in of 0° lay-ups. With reference to Fig. 1, a 90° lay-up is here defined as one where all the fibres are oriented along the length axis of the spar, whereas a 0° lay-up is one where all the fibres are in the plane of the spar cross-section. While one study [9] showed that 0° laminates gave the highest spring-in, other studies [3], [6], [10] found that spring-in of 0° laminates is lower than that of multidirectional laminates.

For symmetrical laminates, the stacking sequence has been found to have little effect on spring-in [6]. The study noted that even when 0° and 90° plies were interchanged, spring-in remained virtually unaffected, provided that the lay-up remained symmetrical. In an experimental and numerical study [10], C-shaped parts were found to have greater spring-in than L-shaped parts. This difference was explained in terms of a possible ‘geometric locking’ of the C-shaped parts on the inner mould line that was not present with the L-shaped parts.

Spring-in due to thermal strain anisotropy is believed to be proportional to the difference between the cure temperature and the ambient temperature, see Eq. (1). However, this is not the only source of spring-in. The effect of cure temperature on the development of residual stresses has been studied. One study found that by processing at a lower temperature for a longer time, or by utilizing an intermediate lower temperature dwell in three step cure cycles, residual stresses can be reduced by as much as 30% [13]. However, processing at lower temperatures requires longer curing times. The same study showed that for a given cure temperature, reducing the cure cycle time can reduce residual stresses, resulting in a decrease of warpage of as much as 60%. However, reducing the cure time does not allow the curing reaction to complete which has a detrimental effect on the mechanical properties. A reduction of 12% in residual curvature was reported for asymmetric laminates when cooled down at a rate of 0.56 °C/min compared to a rate of 5.6 °C/min [13]. This was explained by viscoelastic stress relaxation being more significant for the slower cool-down. The same researchers also reported that varying the cool-down pressure from 0.35 to 1.0 MPa has no noticeable effect on warpage. In contrast, other work has shown that if a 2-hold cure cycle, with a first lower temperature hold and a second higher temperature hold, is designed such that the resin gels during the first temperature hold, spring-in can be significantly increased compared to a single-hold cure cycle [10], [11]. The effect of cure cycle on thickness and fibre volume fraction distribution has been studied experimentally and numerically [24]. The study showed that by optimizing the cure cycle, the quality and strength of the part was enhanced.

A numerical study [12] indicated that reduction of the mechanical interaction between the part and tool at the tool-part interface reduces the spring-in angle. In the same numerical study, the tooling material was predicted to have a major effect on spring-in, with Invar tooling giving lower spring-in than aluminum tooling. Another numerical study [7] showed that the tool coefficient of thermal expansion (CTE) has a direct impact on spring-in. A tool with a CTE of 25 μm/m/°C gave significantly more spring-in than a tool with zero CTE.

The effect of several parameters on spring-in have been reported in the literature [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], as discussed above. However, because of the large number of parameters that vary between studies, it is difficult to draw any general conclusions from this body of work. This paper presents a large amount of experimental data that examines the effect of design and process parameters on spring-in and warpage of composite angles with a symmetrical lay-up. The objective is to determine the effect several design and process parameters have on spring-in, and how the different parameters interact. The paper examines the individual and interactive effects of five design parameters: part shape, lay-up, flange length, part thickness, and part angle, and three process parameters: tool material, tool surface, and cure cycle, on spring-in and warpage of angled composite laminates. The parameters studied were selected based on previous research by the authors [10]. The presented results are based on six sets of experiments, which are fully described in the Appendix. However, only select results are presented in the main text for ease of interpretation. The first experiment was an eight-factor fractional factorial designed experiment. This design allows the examination of eight parameters simultaneously to identify key parameters. The following five experiments were performed to focus on select parameters and to examine interactive effects in more detail.

Section snippets

Calculation of anisotropy spring-in components

Eq. (1) was used to calculate approximate values of the anisotropy spring-in component for parts manufactured in the current study. The parameters used in the calculations and the predicted spring-in angles are shown in Table 1. The material used for all parts in this study was the T-800H/3900–2 carbon/epoxy unidirectional prepreg, which is a non-bleed, toughened system manufactured by the Toray Company. There are several papers in the literature with slightly different elastic and

Tool and part geometries

This section describes the overall range of part and tool geometries used in this study. A more detailed description of each part is found in the Appendix. The tool geometries used in this study are shown in Fig. 2. The tools consist of solid blocks of 6061-T6 aluminum or A36 steel. The tools were machined to an angle of either 90° or 45°, with a corner radius of approximately 6 mm. The tool surface was in a smooth “as-milled” condition. The parts made were either unidirectional laminates [0]n

Results

The measured spring-in of the parts made in experiment 1 is shown in Fig. 7 together with the predicted spring-in, Δθ, using Eq. 1 (Table 1). Although the predicted spring-in is approximate due to uncertainty in the cure shrinkage component, the figure shows that spring-in varies significantly with design and process parameters and that there appear to be mechanisms in addition to anisotropy of cure shrinkage and thermal expansion that contribute to spring-in. For detailed information about

Discussion

Many of the findings in this study are in agreement with previously published work but some are not. For example, the effect of cure cycle found in this study is in agreement with previous work by the authors [10], [11] but in disagreement with other published work [13]. The different trends observed are likely caused by different mechanisms, which are dominant at different process conditions.

Warpage of flat sections was found to be of significance in this study. The results indicate that

Conclusions

The objective of this work was to determine the effect of design and process parameters on spring-in of angled thermoset laminates. The main conclusions from the current work are as follows:

  • Spring-in varies greatly as a function of design and process parameters for the parts made in the current study.

  • For practical reasons, spring-in is usually measured based on the deflection of the flanges away from the corner. The current work showed that spring-in measured this way is sensitive to flange/web

Acknowledgements

The authors would like to thank Dr. Karl Nelson and Mr. Kurtis Willden of The Boeing Company in Seattle for providing materials, Mr. Roger Bennett and Mr. Ross McLeod of The Department of Metals and Materials Engineering at UBC for their technical support, and the Natural Sciences and Engineering Research Council of Canada for financial support of the research. Finally we would like to thank past and current members of the UBC Composites group, in particular Dr. Anoush Poursartip, for

References (25)

  • A. Stephan et al.

    On the springback effect of CFRP stringers: an experimental, analytical and numerical analysis

    Proceedings of the 28th International SAMPE Technical Conference

    (1996)
  • Fernlund G, Poursartip A. the effect of tooling material, cure cycle, and tool surface finish on spring-in of autoclave...
  • Cited by (266)

    View all citing articles on Scopus
    View full text