This article also appears in
Subscribe now »

The final demonstrator structure for the research project was a liquid-resin-infusion (LRI) wing bow spar. Simulation shows filling times for the spar.

Infusing lightweight composite structures

Liquid resin infusion (LRI) is a proven manufacturing technology for both small- and large-scale structures for which, in most cases, experience and limited prototype experimentation is sufficient to get a satisfactory design. However, large-scale aerospace and other vehicle structures require reproducible, high-quality, defect-free parts with excellent mechanical performance. These requirements necessitate precise control and knowledge of the preforming (draping and manufacture of the composite fabric preforms), their assembly, and the resin infusion.

The INFUCOMP project is a multi-disciplinary research project to develop necessary CAE tools for all stages of the LRI manufacturing process. An ambitious set of developments has been undertaken that build on existing capabilities of leading drape and infusion simulation codes available today. The main objective of the European Research Consortium is to develop liquid composites molding (LCM) for the aeronautic sector.

Currently the codes are only accurate for simple drape problems and infusion analysis of resin transfer molding (RTM) parts using matched metal molds. Furthermore, full chaining of the CAE solution will allow results from materials modeling, drape, assembly, infusion, and final part mechanical performance to be used in subsequent analyses.

Although the materials and manufacturing methods in INFUCOMP are specific to aerospace structures, the research work is expected to be of great value to other industries, including automotive and manufacturing, energy, rail, and marine.

INFUCOMP has built on PAM-RTM, an existing simulation software from ESI Group, to provide a full solution chain for LRI composites, including fabric modeling, drape, assembly, infusion, cost, and final part performance prediction. Simulation tools will avoid costly and time-consuming prototype testing, allow the CAE design of alternative manufacturing routes, and enable cost-effective, efficient LRI composite structures to be designed and manufactured.

During the industrial validation phase of the project on simplified components and a relevant LRI aircraft substructure, researchers from ESI Group, Daher Aerospace, University of Stuttgart, and INASCO (Integrated Aerospace Sciences Corp.) employed numerous enhancements to the state-of-the-art for resin infusion simulation—in particular, better viscosity models and essential developments to run under distributed memory processing (DMP) to take advantage of new-generation cluster computers and massive parallel computing. This includes coupling of modeling and monitoring, allowing a combination of predictive capabilities provided by simulation with the capability of detecting unexpected events and variations in real time provided by process monitoring.

Experiment setup

For the project, researchers employed the DiAMon Flow monitoring system developed by INASCO, which combines flow, thermal, and cure monitoring as part of an integrated sensor comprising electric contact pairs, an inter-digitized micro capacitor, and a thermocouple. Capabilities include measurement of the in-plane position and speed of the flow front, estimation of the degree of cure (through material state models that result from the correlation between the dielectric measurements and available cure kinetics models), and continuous monitoring of temperature.

All the sensors used in the system are resistance sensors: When the resin arrives on the sensor, a 10 V dc passes from one of the pins through the resin and into the other pin. From the voltage drop that is recorded by the “high resistance sensing electronics,” the presence of resin on this location, the value of the resin resistance, the arrival time of the resin, and the time since the resin arrival are noted. The sensing area of each individual sensor is protected from direct contact with the carbon fabrics to prevent the sensor from short-circuiting. This protection uses two layers of glass cloth positioned on top of the sensing area.

The resin used in this project is RTM6 from Hexcel. The preforms used are made of Hexcel twill fabric 48302 weaved with carbon fibers T700 12K.

Material data were taken from previously conducted experimental and modeling work:

• Resin viscosity: 0.033 Pa.s-1 (taken from RTM6 manual Hexcel); also, a new constitutive model has been applied

• Distribution medium permeability (measured by University of Stuttgart): K=0.74E-10 m²

• Preform (48302 T700 Hexcel reinforcements) having in-plane permeabilities K1=K2=4.48E-12 m² and through thickness permeability K3=2.79E-14 m².

The distribution medium is a two-dimensional flowing aid used to ease resin flow and distribute the resin across the surface of the laminate. It is commonly accepted to consider K1=K2=K3 for distribution media.

Validations on simplified components

Validations on small-scale singularities that are representative of industrial issues were conducted. For singularity #1, “step,” the part to be infused has two preforms: a lower preform of 3 mm (0.12 in) thick and a thicker upper preform of 4.2 mm (0.165 in). The sensors were positioned, and simulations were carried out with PAM-RTM 2013. A 549,000 tetrahedron mesh was generated with Visual-Mesh. This mesh was split into two zones, with one zone for the distribution medium and one zone for the preform. Distribution medium permeability tensor was taken: K1 = K2 = K3 = 2E-10 m² (calibrated to match with experiment).

Simulation results show that resin flows preferentially into the distribution medium; however, the preform starts to be impregnated before the complete filling of the distribution medium, resulting in an orthotropic flow process.

Experimental measurements give a filling time of approximately 900 s (15 min). The distribution medium permeability tensor used in simulation has been calibrated to match with the experiment result. This flow media permeability tensor will then be used for all other cases to validate the simulation tool.

For singularity #2, “Ω stringer,” the manufacturing part is made with a simple preform that is 3 mm thick, 280 mm (11.0 in) wide, 380 mm (15.0 in) long, and reinforced by a Ω stringer that is 3 mm thick and 93 mm (3.7 in) wide. Sensors were positioned, and a 79,600 tetrahedron mesh was generated with Visual-Mesh. The mesh was split into three zones, and material data were taken:

• Resin viscosity: 0.033 Pa.s-1 (taken from RTM6 manual Hexcel)

• Distribution medium permeability: K1=K2=K3 = 2E-10 m² (calibrated for the step case)

• Preform (48302 T700 Hexcel reinforcements) permeability (measured by Hexcel reinforcement):

• K1=K2=4.48E-12 m² in the plate and in the Ω

• K3=2.35E-14 m² in the plate (vf = 58%) and K3=1.974E-14 m² in the Ω (vf = 61%).

Experimentally, resin passed through the vent at 24 min (1440 s) after the start of infusion. The vent was clamped at 45 min (2700 s) after beginning of infusion. According to monitoring results, the resin maintained progress after clamping and triggered the sensors at 60 min (3600 s).

The simulation results show a resin arrival at vent location after 1447 s and a total filling time of 3656 s. These results are extremely close to the experimental results.

Industrial demonstrator

The final demonstrator is a generic wing bow spar—an industrially relevant LRI aircraft substructure. Distribution media has been located on both sides of the part based on simulation results to optimize through thickness flow and filling time. The final demonstrator was equipped with several DiAMon Flow sensors to compare experimental results with simulation results.

A 950,424 tetrahedron mesh was generated with ESI’s Visual-Mesh software. Four zones (group of elements) were generated to define materials properties. One for the two distribution media, one for the gusset filler, one for the two “C” frames forming the “I” stringer, and one for the top and bottom plates.

The preform is made of the woven fabric type 48302E01 and the unidirectional fabric type X505E01 provided by Hexcel reinforcements, and permeabilities of those two reinforcements were measured by Hexcel reinforcements.

• 48302E01: K1 = 9.76E-12 m²; K2 = 9.76E-12 m²; and K3 = 1.13E-11*exp(-9.8 vf) m² (vf = fiber volume fraction)

• X505E01: K1 = 2.03E-12 m²; K2 = 2.03E-12 m²; and K3 = 1.11E-12*exp(-9.82 vf) m².

The bottom and the upper plates are made of both 48302E01 and X505E01. To represent properly the preform permeabilities, an equivalent permeability was computed for the whole stacking. The “C” frames constituting the “I” stringer are made only with the 48302E01 reinforcement at 57% of fibers.

Permeabilities used in the simulation are:

• Bottom/upper plates: K1 = 3.68E-12 m²; K2 = 3.68E-12 m²; and K3 = 1.5E-14 m²

• “C” frames: K1 = 9.76E-12 m²; K2 = 9.76E-12 m²; and K3 = 4.24E-14 m².

To model the resin inlet, the researchers used a flow rate boundary condition with a maximum pressure. As long as the inlet pressure is below the maximum pressure, constant flow rate is imposed at the inlet and constant pressure equal to maximum pressure is imposed. This kind of boundary condition allows the proper description of pressure evolution at the inlet without modelling the full injection line from the resin pot to the mold.

Three injection points (inlets) were defined on each channel used in the real experiment by Daher. Numerically, 1.11E-6 m3/s was imposed on each of those inlets with a maximum pressure of 1E5 Pa (= 1 bar). Two vents were defined numerically and were set to 0 Pa.

The current framework for viscosity modelling in the context of composites processing is based on the use of temperature and degree of cure or glass transition temperature as state variables. This implies that a viscosity development model needs to be coupled with a model of the cure kinetics, while the implementation of the model requires a series of cure kinetics characterization experiments in addition to the necessary rheological tests. This type of model works well in the context of autoclaving where the value of viscosity is important over a wide range of degrees of cure (from the uncured material up to gelation).

However, the role of a rheological model in the simulation of liquid molding processes is different, as viscosity is one of the parameters governing the filling/infusion stage during which the changes in degree of cure are relatively small. These small changes induce a significant increase in viscosity, which can eventually alter the outcome of the process. When models based on the use of the degree of cure are utilized, the accuracy of the simulation can be compromised by the fact that one of the underlying variables of the model has a small variation during the process. This can result in usage of a model that is developed over a range of wide degrees of cure (from about 0 to 60%) only within a limited range (up to about 10%).

The approach adopted by INFUCOMP researchers overcomes this limitation by using the viscosity at a reference temperature as a state variable instead of the degree of cure. The reference viscosity follows its own kinetics.

Due to the higher permeability of the distribution medium, resin preferentially flows in this flow media. This phenomenon favors a transverse flow in the central part of the spar. Finally, infusion time (2157 s) is in good correlation with experimental results (2100 s). Moreover, the last-filled zones correspond with good precision to dry spots identified on the actual part.

Again, simulation results were in agreement with experiments and found to be valuable to understand and validate the infusion process. The results also help researchers to understand the simulation workflow and methodology as well as the monitoring capabilities.

This article is based on SAE International technical paper 2014-01-0965 written by Pierre Marquette and Arnaud Dereims of ESI Group; Michael Hugon and Guenael Esnault of Daher Aerospace; Anthony Pickett of the University of Stuttgart; and Dimitrios Karagiannis and Apostolos Gkinosatis of INASCO.

Continue reading »