Inside Analysis: Simulating VARTM for better infusion

A resin flow simulation tool is helping to optimize liquid molding of aerospace parts.
#outofautoclave #layup


Facebook Share Icon LinkedIn Share Icon Twitter Share Icon Share by EMail icon Print Icon

The progression of resin through a mold during infusion is a complex and often inscrutable phenomenon. Seat-of-the-pants mold design based on past experience might work for simple part geometries, but as shape complexity increases it becomes more difficult to position mold vents and gates for the best results. Getting it wrong can mean incomplete wetout of the reinforcements, nonuniform cure, poor resin/fiber bond and residual stresses.

“There is a real need for analytical tools to supplement experience in manufacturing design,” says Dr. Jeffrey M. Lawrence of V System Composites Inc. (VSC, Chester, Pa. and Anaheim, Calif.), previously a staff member at the University of Delaware’s Center for Composite Materials (UD-CCM). UD-CCM’s Drs. Suresh Advani and Pavel Simacek have, for the past 12 years, developed the Liquid Injection Molding Simulation (LIMS) tool, which enables the user to digitally model and predict a resin’s in-mold flow behavior in resin transfer molding (RTM) and vacuum-assisted resin transfer molding (VARTM) processes. The LIMS software, to which Lawrence contributed while at the University, has become a key tool in VSC’s efforts to optimize tool, part and manufacturing process designs.

In addition to the simulation model, the V System team also uses a resin flux process monitoring method, which allows the infusion process to be monitored in real time. Together, the modeling and monitoring ensure high-quality repeatability for better parts, says Lawrence.

LIMS helps part transition

VSC employed LIMS to help validate VARTM as an alternative process for fabricating a complex aerospace part, one that HPC has covered previously (see “Learn More,” p. 41). VSC replaced a multipart metallic design with a single, infused carbon/epoxy part that included monolithic laminate blade stiffeners arranged in a grid pattern, supporting an integral monolithic skin, co-infused and cocured as a single part (see Step 1 of the process step photos, which can be viewed on p. 3). LIMS played a key role in helping the company demonstrate the manufacturability, performance and affordability of infusion for high-performance aerospace parts.

Just as structural analysis has helped avoid overdesigned parts and excessive safety factors, the LIMS tool is intended to build robustness into the infusion process, explains Lawrence: “Closed mold processing has certainly helped health and safety by reducing volatiles in the shop, but the downside is the inability to see the resin flow.” The only clues to resin progression are resin ingress into the gate and egress out of the tool vents. Fabricators, then, could assume infusion is complete and open the mold only to find dry areas in the preform.

With LIMS, analytical techniques are applied to the infusion process, enabling simulation of resin flow through a preform in a virtual trial-and-error process. Lawrence explains that the program’s flexibility allows the user to interrogate the state of mold filling or change the boundary conditions at any step, essentially enabling simulation of the opening or closing of any gates or vents and pressure variation. “This makes it feasible to test scenarios, such as sequential injection, or control strategies for optimum results.”

The governing equations used in LIMS for resin infusion are the two-dimensional and three-dimensional velocity/pressure relationships of Darcy’s Law, based on the physics of flow through an anisotropic porous medium that correlates the instantaneous discharge rate, the viscosity of the fluid and the pressure drop over a given distance. For RTM, two-dimensional flow is assumed because the pressurized resin front is contained with fixed tooling (see Step 2). VARTM, explains Lawrence, often becomes a three-dimensional problem because the layups are generally thicker and contained under a flexible vacuum bag — resin has to flow horizontally and vertically through the distribution medium to wet out the preform. The equations are solved numerically, using a finite element/control volume approach. Says Lawrence, “While each fluid flow path would be impossible to track individually, the volume-averaging approach posed in Darcy’s Law allows for macroscopic description of the resin flow progression for greater simulation power.”

The first step in the use of LIMS is the generation of a mesh. LIMS has no inherent meshing capability, so another software tool, in this case, ABAQUS supplied by SIMULIA (formerly Abaqus Inc., Pawtucket, R.I.), was used to create a mesh from the solid model shown in Step 1. The mesh itself is shown in Step 3. Linear triangular elements connect 10,703 nodes for the example aerospace part.

Next, says Lawrence, the team added to the model the resin runners that the company designs into its trademarked HypPerVARTM tools. HypPerVARTM does not require any separate infusion media but rather incorporates a system of resin distribution runners and channels. These enable rapid in-plane and out-of-plane wetout of the reinforcements. One-dimensional model elements (shown in yellow in Step 4) were added along the resin runners to represent the resin flow channels, explains Lawrence. Their permeability is assumed to be higher than the dry bulk fabric, with permeability equivalent to flow through a duct.

For the simulation, resin injection “gates” were applied as boundary conditions at 10 locations. These consist of single resin injection points that feed the resin distribution runners, says VSC’s president Mike Louderback (the company reportedly is close to receiving several patents for this system). The results of the simulation are shown in Step 5. “The resin initially fills the 1-D elements, representing the flow channels, and then begins to saturate the neighboring skins of the part, radiating outward from the center point. Resin then progresses up the ribs and, finally, down the lower legs of the part to complete the infusion,” reports Lawrence. “The simulation suggested that the resin progression would completely wet out the preform, ending up moving down the lower legs of the part towards the ultimate vents.”

Virtual vs. the real world

In the case of the blade-stiffened aerospace part, the design of the resin distribution system seemed ideal based on the simulation, so tooling was fabricated to produce a demonstrator part. The company made its own carbon/epoxy outer mold line (OML) female tool (so named because the tool’s inside surface forms the finished outer surface of the part). To form the blade stiffeners/ribs, nine aluminum blocks were machined by an outside supplier to fit into the mold cavity as inserts, with sufficient tolerances to accommodate the required carbon fabric plies. After initial layup of the skin plies, carbon fabric was wrapped around five sides of each of the aluminum blocks to form the vertical blades (leaving the top surface uncovered so that the blocks can be removed after cure); the wrapped blocks then were fitted together in the mold to unitize the structure, says Lawrence. The inserts function to place consolidation pressure on the vertical blades as they expand during heat-accelerated cure because the aluminum has a higher coefficient of thermal expansion (CTE) than the composite tool itself (Step 6). Bleeder material was placed over the layup, then the entire layup and tool was vacuum bagged and placed in an oven for infusion and cure (Steps 7 and 8).

To verify and validate the model, it was important to monitor the VARTM process and the amount of resin flowing into and out of the layup — the “resin flux.” Notes Lawrence, “While simple in concept, the process monitoring method was an essential element of an intelligent manufacturing approach and a complement to the simulation provided by LIMS.”

VSC tracked how much resin was introduced into the part and how much exited through vents via resin “bleeding” by weighing the infusion bucket and the vent bucket with digital scales connected to a computer (see Step 9, on p. 3). By observing the time rate change of the resin mass in the infusion bucket, the flow rate into the mold could be calculated. Doing the same with the vent bucket enabled calculation of the difference between the two, which represents the resin flux or mass flow progression through the mold.

“By comparing this actual measured data over time to the simulation, one can ensure that the infusion is progressing as expected,” states Lawrence. “If it is flowing too slowly, a blockage may be present — perhaps plies that are askew or some other type of flow restriction. It gives you the opportunity to find the blockage and remedy it before any adverse effects are created.” Another benefit of this process-monitoring step is that part-to-part variation can be studied in the production environment in support of quality assurance efforts during manufacturing. The monitoring program was written in a graphical programming language called LabVIEW, a software program geared toward data acquisition and supplied by National Instruments Corp. (Austin, Texas).

Although they were not used for this demonstration aerospace part, Louderback points out that resin flow sensors also can be incorporated into VARTM tools and used in concert with LIMS and process monitoring to gain additional information on local resin behavior in complex tool geometries and preform materials. Although resin bleeding during VARTM, after resin has “witnessed” or flowed out of the vents, is a common practice to ensure that any remaining air in the reinforcements has been expelled, Lawrence predicts that, “as LIMS is used more widely and with more confidence, this wasteful practice will become unnecessary,” and he adds, “the program is continually being updated with new features and to accommodate newer processing approaches for more precise results.”

For the demonstration aerospace part, the actual mold-fill data were compared to the LIMS model, and the results showed close agreement, says Lawrence (Step 10). The deviation between the two plots is mainly due to some inaccuracy in the estimates made for the permeability of the fiber reinforcements, as well as simplifications in the modeling of the part geometry. The waviness in the experimental data near the end of the infusion is due to the need for repeated emptying of the resin bleeding catch bucket.

The company believes the comparison demonstrates that VARTM is sufficiently mature to be implemented in aerospace manufacturing programs. “The key benefit to being able to simulate the VARTM process is that an entire part, despite its complexity, can be successfully tooled and infused in one shot, instead of having to secondarily bond features such as stiffening ribs, saving time and labor,” concludes Lawrence.