Evaluation of Laminate composite Distortion by an Integrated Numerical-Experimental approach

Numerical Model - Phase 1

Preliminary work

Selection of a thermo-chemical model which defines the heat transfer between the part and the tool/the environment.

Definition of the thermochemical model

Selection of a cure kinetic model which predicts the evolution of the cure rate.

Definition of the cure kinetics model

Selection of a cure shrinkage model which defines the volume shrinkage during cure.

Definition of the cure shrinkage strain model

Selection of a non-linear model which simulates the resin modulus variation versus temperature

Definition of non-linear elastic model

Utilization of the homogenization technique in order to achieve the composite properties at various moments of time in the curing process

Composite properties at various temperatures obtained through homogenization technique

Numerical simulation - First results achieved

Distortion on Z axis in the skin coupon

Displacements in the coupon in the initial simulation amplified

Displacements in the C-shaped spar coupons

Displacements in the C-shaped spar coupons

Observation: The uncalibrated numerical tool has provided inaccurate distortion results. A coupling reaction can be observed for the skin coupon as for the C-shaped coupon, a warpage effect can be observed combined with spring-out of the flanges.

Calibration of the numerical tool

For the calibration of the numerical tool the following steps have been completed:
      - Insertion of the resin gel point as parameter in the simulation process according to the AIMEN experimental results
      - Insertion of a variable friction coefficient. Friction between the part and the tool starting with the resin gel point. No friction below gel point.
      - Insertion of a variable through-thickness fiber volume fraction which takes into account the uneven distribution of resin in the LRI (Liquid Resin Infusion) process in the C-shaped spar coupon and the maintaining of the constant fiber volume fraction for the skin coupon.

Displacements distribution in the skin coupon after calibration

Displacements distribution in the C-shaped spar coupon after calibration

FE plot of C-shaped coupon spring-in. The distorted part on top of the tooling

Experimental Approach - Phase 1

COUPON TESTING: Monitoring Manufacturing and Shape Distortion Analysis

Liquid resin infusion

Small scale coupons were manufactured by liquid resin infusion of RTM6 resin and dry carbon fiber preforms manufactured by AFP.

Steps followed in the experimental approach:

Monitoring manufacturing results

The manufacturing process was monitored with Direct Current (DC) sensors and Fiber Bragg Gratings (FBG). DC sensors indirectly measure resin cure behaviour through resistivity which changes during the curing reaction with the formation of polymeric chains and crosslinking. DC sensors measure temperature and electrical resistance of the resin. These parameters can be related to the resin behaviour during manufacturing. DC sensors were able to detect resin arrival, point of minimum viscosity, gel point and end of cure during the manufacturing cycle.

FBG sensors measure temperature and strain by means of a shift in a designed wavelength reflected. FBG temperature sensors were placed at both surfaces of the coupon detected the temperature distribution across the surface of the coupon and the through the thickness of the laminate.

FBG strain sensors detected a significant variation of strains (contraction) due to resin cure shrinkage at the gel point and more pronounced decreased of strain during the cooling stages. At the end of the curing cycle an uneven distribution of strains was observed across the coupon.

Post-manufacturing analysis of shape distortions

3DCMM measurements were performed after demoulding during several days monitoring changes on the final part after manufacturing. The same FBG sensors used during manufacturing were reconnected and continue monitoring strain evolution after demoulding.

Skin coupons after 14 days of manufacturing showed an increased in the curvature along the X axis. C-spar coupons showed warpage of the web and spring-in of the flanges. The strain sensors effectively monitored release of strains after the day of demoulding.


The same methodology was used for monitoring prepreg cure evolution, temperature profile and strain variations during the manufacturing cycle. After fabrication the geometry and strain evolution were also analysed in the same way as for LRI coupons. In this case, the sensors were also embedded in between layers to get a better understanding of the through-thickness variation of temperature and strain.

Embedment of sensors on prepreg skin coupon and C-spar coupon.

Numerical Model - Phase 2

A sub-section of a full wing-box assembly has been modelled. The part consists of three spars and skin. The tooling consists of the skin tool, the spar tools, fixtures and auxiliaries.

Downscale demonstrator Tooling Assembly

The FE model of the entire assembly needs a certain degree of simplification.

Simplified geometry of the downscale demonstrator

FE model of the tooling and part assembly

The meshed parts assembly manufactured in the 1.2m downscale demonstrator

Close-up view of the FE for the C-spar-Skin junction

Close-up view of the FE for the I-spar-Skin junction

The meshed parts assembly positioned on the Skin Tool (red)

In order to obtain a robust model that does not require high computing resources, the Skin Tool substructure and Spar Tools have not been modelled using 3D elements. Instead, constraints that bear the same role have been applied.

The fixture and Skin Tool substructure were replaced by constraints on the Z direction

Constraints applied to simulate the effect of the Spar Tools


Degree of cure at the end of the curing process

Observation: At the end of the oven curing cycle, the degree of cure is not homogeneous but has a certain variation throughout the part, as described in the picture. The complete curing is obtain after a post-curing phase.

Z direction displacements [mm]

Distorted shape of the part amplified by 30 times

Conclusions: The first iteration of results is showing geometrical deformations of the spars that are similar to the spar coupons which have been simulated in the first numerical phase. The spars show spring-in of the flanges in the range of 0.3mm up to 2.18mm [Z displacement]. The skin shows different deformations compared with the skin coupons. A calibration phase is required in order to achieve a both accuracy and precision for the simulation.

Numerical Model - Phase 3

Phase 3 consist in developing a numerical model for the entire wing box assembly. It was modeled the part (three spars and one skin), the Spar Tools, the Skin Tool.

The full-scale Part and Tooling Assembly

Detailed FE model of the full-scale demonstrator

Because of the geometry complexity, the FE model of the entire assembly needs a certain degree of simplification:
       - The back structure of the support intended to be replaced with finite element constraints
       - The bridges also intended to be replaced with equivalent finite element constraints

Detailed on mesh

In the same was as presented in previous phase, in Phase 3 were created two different models for the two different manufacturing techniques used for the wing box:
     - Liquid Resin Infusion (LRI) manufacturing method
     - Pre-preg manufacturing method

Phase 3 results:

LRI part distorsion

Prepreg part distorsion

In order to make a comparison between the distortion occurred at the two methods described above, and to isolate the spring-in and spring-out effects in the wing skin we used CIRCULAR regression functions to compute a mean curvature radius of the outer surface of the skin.

A spring index was defined

SI = (Mean Radius of the Undistorted Section/ Mean Radius of the Distorted Section -1) * 100

If       SI > 0, Spring-in effect
Else   SI < 0, Spring-out effect

Several sections, with different material thickness were chosen to compute the Spring Index:

Sections positions for spring in/out study

Spring in/out SI comparison for LRI

Spring Index comparison between LRI and prepreg method

Also, the direct method measuring was used to compare the distorted part with the undistorted reference part.

Conclusions: Both simulations of LRI and prepreg had a similar behavior that can be compared with the reality. The final degree of cure is similar with the one resulted from the experimental tests. For the tuning of degree of cure the pre-exponential factor can be utilized.