Modeling sand core blowing: Simulation's next challenge. (Technology in Progress).

Since the '80s, foundries have used computational fluid dynamic (solidification) software to predict metal flow in the mold. Foundries run these programs with casting design models to optimize riser design and to study shrinkage, microstructures, stresses and other casting properties to ensure quality components.

While this software continues to develop and increase in ability, another area for fluid dynamic modeling is being explored--sand core blowing and gassing. In the past, success with modeling core production hasn't been achieved due to the complexity of the fluid/granular process. However, due to experimental work performed by General Motors Powertrain (GMPT) and Arena-flow, L.L.C. on a production engine block water jacket and slab corebox, modeling sand core blowing has become a reality.

This article takes a look at the research performed at GMPT to validate the new modeling system for coldbox sand core blowing.

Core Modeling Software Overview

Similar to metal solidification software, the sand core modeling software is a 3D math-based computer code. However, the core blowing system is tailored for particle flows where the characteristics of individual particles are important, such as in sand or lost foam bead flows. The numerical method provides an economical solution for two-phase flows, possessing a virtually unlimited range of particle types, sizes, densities and shapes. It calculates particle flows ranging from a dilute mixture to the close-pack limit.

This computational fluid dynamics (CFD) approach uses multiphase particle-in-cell (MP-PIG) numerical methodology. A new innovative Eulerian-Lagrangian algorithm models the air as a fluid and sand grains or polystyrene beads as both discrete particles and as a fluid. In a Lagrangian frame of reference, particles are not tied to a grid, but are free to move in space from forces applied to each particle. The number of particles modeled in this water jacket calculation exceeded 500,000.

In pure Lagrangian flow, forces such as particle-to-particle collisions are difficult to calculate, whereas in Eulerian flow (grid), these forces are handled well by gradients. In the MP-PIC scheme, the particles are mapped to the grid and treated as a fluid to calculate particle spatial gradients. The gradients then are mapped back to particle locations to compute the forces on each particle. This new method uses the strong points of both numerical approaches.

The math-model and numerical scheme allows particle flow calculations from dilute particles to close-pack, where close-pack is the maximum packing of a granular material (typically on the order of 50% to 70% volume of particles). Because particles are not simply tracers flowing with the air, the modeling of forces and properties of discrete particles provides the ability to calculate subgrid characteristics such as air-to-particle drag, wall impaction and erosion, distinct material interfaces and local density, unfilled or poorly filled regions, distribution of particle sizes in regions and lost foam bead expansion at hot walls.

Validation at GMPT

GMPT's Manufacturing Analysis Group worked with the Defiance, Ohio-based GMPT foundry to validate the core modeling software on a production core blowing machine. In contrast to a laboratory setting where instrumentation fits the experiment and testing is flexible, measuring data in a production process is restrictive, The measurements were limited to transient pressure data and post-blowing core examinations.

The corebox chosen for experimentation and analysis produces three water jacket cores and three slab cores per cycle (Fig. 1). Defiance patternmakers "fitted" a jacket core cavity to allow pressure measurements while maintaining the proper seal during the blow cycle.

Experiments were run under normal machine settings and with variations in blow pressure. In each case, specified blowtubes and/or vents were blocked. Pressure was measured in the accumulator air supply-line next to the magazine and in the magazine--40 cm up from the blowplate. Two pressures were measured on the blowplate--one measured 9.2 cm above the blowplate and the other at the entrance to a blowtube. Three pressures were measured in the core cavity, two at the exit of blowtubes and one at the top wall.

GMPT's Measured Production Data

When the measured data (Fig. 2) was first examined, it was perplexing, One immediate observation was that filling of the sand cores is fast as it takes approximately 0.7 sec to fill the three jackets and three slabs.

The relative response between measured pressures appeared to be inconsistent. For example, the accumulator feed-line pressure steps to 460 kPa (66.7 psig) while the pressure in the core remains unchanged at near atmospheric pressure for the first 30% of the fill period. When the corebox pressure does rise, it only rises to a modest 60 kPa (8.7 psig) and remains nearly constant. A question arose as to when air and sand actually began flowing into the corebox, and why did the cavity pressure remain so low. There also was the question of what caused the pressure "spikes" or "bumps" that occurred only in the core cavity. From the measured data alone, there was no good explanation for the data behavior, let alone what was happening in the process.

Simulation Model Data

A simulation model for the water jacket and slab cores was built using a CAD solid model. The meshed model comprising the magazine, blowplate and one core is shown in Fig. 3.

Prior to the research discussed in this article, fundamental experiments were conducted by GMPT to obtain the sand grain size distribution, sand-covered vent behavior, and drag factor between air and binder-coated sand. The boundary condition at the top of the magazine was the measured magazine pressure, and the core was vented to the atmosphere through 41 slot and screen vents, each discretely represented in the model. The core blowing simulation took less than one day of computer time.

The calculated and measured pressures compared well for timing of events as well as the shape and magnitude of the pressure trace, As a result, it was reasoned that for the computed (simulated) pressure to compare so well with measured (production) data, then the overall core filling behavior must have been calculated very well. This rationale allows reasonable explanations to be formed for the observed pressure behavior based on results from the math-based modeling tool.

Core Blowing Modeling

Figure 4 shows time snapshots of the calculated core blowing.

The blowing process begins with discharge of air from an accumulator that chokes where the air enters the magazine. Choked flow lasts approximately 0.1 sec until the magazine pressurizes. Sand and air then immediately flow into the core. However, air flowing in the core easily exhausts through the large number of vents, and the core pressure remains near atmospheric.

This effect is responsible for the measured "delay" in the core cavity pressurization.

As the core fills, the lower vents cover with sand, partially blocking the escaping air, and the core begins pressurizing at 0.25 sec, which is seen in both measured and calculated data. But why does the core cavity pressure remain 8-10 times lower than the driving accumulator air pressure? The answer is that air driving into and through the sand filled magazine has substantial pressure loss from sand-to-air drag. Further, air is channeled into the small diameter blowtubes producing high air velocities and a large pressure loss. The large drag on air flowing through the sand provides the major pressure loss in the system, and the core cavity pressure only rises to 8 psi.

Another question: Why does the core cavity pressure remain constant while the blowplate and magazine pressures increase, even after the corebox is filled? This behavior is observed in the measured and calculated pressure histories. Examination of the model's computed pressure distribution throughout the entire machine reveals that a large pressure-drop occurs along the blowtubes after filling. This occurs as high-speed air flows clown the sand-filled blowtubes.

Increasing the magazine pressure only causes a higher velocity and pressure drop in the blowtubes with a small increase in the core cavity pressure. Calculated pressure distribution throughout the blowing machine confirms the blowtubes cause the principal pressure drop.

An unexpected and interesting behavior is "spikes" or bumps in both the measured and calculated pressure data near the bottom of blowtubes. The pressure spikes were not seen in the magazine or blowplate pressures, indicating they were associated with an event in the core cavity.

The probable cause for the pressure spike was the final filling of the core tooling due to the stopping of particle flow and "squeezing" of air as sand finally fills that section of core immediately under a blowtube. The modeling software calculates the pressure spikes to occur just as sand fills this region, quickly forcing air into the surrounding sand. The modeling of sand filling occurs in a 3-D manner with last-to-fill regions being underneath the blowtubes, causing the spikes seen in all experimental data.

Future Sand Core Modeling

The immediate future for sand core simulation is the gassing and core hardening process. These activities will further the metal casting industry goal of using advanced math-modeling to improve casting quality and economics.

In 1999, GMPT provided the initial funding for development of the modeling. In 2000, NASA (administered by Auburn Univ's SDCC) provided additional R&D funds. In 2001, the U.S. Dept. of Energy matched funds with partners GM, Vulcan Engineering, Styrochem, Foseco-Morval and Mercury Marine to extend the CED technology to lost foam pattern forming.

For a free copy of this article circle No. 344 on the Reader Action Card.

[Figure 2 omitted]

RELATE ARTICLE: What Is a Math-Based Modeling Tool?

Today, foundrymen are accustomed to running casting modeling software programs to answer a variety of questions. As with an automobile, however, many people effectively use it without understanding what is under the hood.

Casting modeling computer programs are math-based tools. A math-based tool solves equations that model physical phenomena important to the process. Fluid flow programs, such as casting simulation software, are based on Newtonian physics [F=ma (force equals mass times acceleration)].

Complexity in the simple Newtonian model arises when applied to a continuum (fluid) with a multitude of driving forces and variable physical properties within complex geometry. Further, because the fluid dynamic equations are nonlinear, the governing equations must be solved numerically where the problem space is divided into finite cells.

To understand how a math-based tool can be valuable, it must first be recognized that all "math-based" models are not perfect. Only the phenomena itself can provide the exact picture. A model only is an approximate representation of physical phenomena. The correct question to ask is whether the model is useful.

For a model to be useful, it must have an acceptable level of accuracy. Accuracy has two measures. First, how well does the math describe the phenomena being analyzed? Second, the math-based tool must have acceptable accuracy in predicting the behavior being studied. This verification of accuracy comes through comparing calculations with experiments.

About the Authors

Ken Williams is a Registered Professional Engineer specializing in modeling of two-phase fluid flow. Dale Snider is a mechanical engineer specializing in computational fluid dynamics. Mike Walker and Sheila Palczewski are senior manufacturing engineers in the manufacturing design, simulation and analysis department at GMPT.

For More Information

The authors of this article will be part of a presentation, "Process Modeling: Sand Core Blowing," at 3:45 p.m. on May 6 at the 2002 AFS Casting Congress in Kansas City, Missouri.

"Process Variables in Core Blowing," Foundry Trade Journal, October 1993.