A rapid alternative to solidification modeling.

Driven by a PC, cellular automaton software locates areas prone to defects in a fraction of the time required by other methods.

To lower costs and reduce expensive lead times on new components, progressive foundries are beginning to implement computer simulations of casting solidification using new software packages.

Modeling through computer solidification can revolutionize design methodologies while addressing the economic needs.

This article describes the MAVIS cellular automaton software package. Developed by the authors, the system can rapidly simulate the macrofreezing pattern of sand and die castings on a standard personal computer (PC). Both ferrous and nonferrous foundries in the U.K. use the system to design and/or modify methoding systems to produce castings free from macroshrinkage defects at maximum yield.

Current Programs

Over the last decade, many computer programs have been developed to aid casting design. These software programs are either commercially available or were developed by companies in-house to allow computer modeling of various aspects of casting processes.

Commercial packages differ in their design function, ease of implementation into current working practices, operation principles and cost.

The programming techniques underlying the various packages range from CPU-intensive numerical simulations of the approximate physics of the process (such as finite element method) to empirical- and/or knowledge-based systems (such as modulus methods/expert systems).

Generally, simulation is used to obtain progressive freezing of the casting toward the feeders, thus avoiding isolation of pockets of solidifying material and the resulting shrinkage defects. If this can be achieved, the feeders are then made as small as possible to reduce wastage and maximize casting yield.

Numerical modeling techniques attempt to solve the real physical equations describing different aspects of casting process through heat transfer, solidification, fluid flow and solid-state deformation.

However, the accuracy of such models relies on the mathematical accuracy with which the various physical phenomena can be described and the accuracy of the numerical procedures employed to solve the mathematical expressions. To use a numerical model effectively, it requires:

* accurate thermophysical materials properties data;

* all boundary conditions and their associated heat transfer coefficients (such as metal/mold, die/air, die/cooling channel);

* a metallurgically sound description of freezing and the related latent heat release.

In addition, different alloy systems may require different freezing models, depending on the particular aims of the simulation.

Insufficient precision in any of these areas may significantly reduce the accuracy of the resulting simulation. Numerical models, by their very nature, may require powerful computer hardware, a high level of user expertise and prohibitively long simulation times (days or weeks).

Alternative Method

While much effort is being expended worldwide on the improvement of numerical models, there have also been developments in software that avoid solving physical equations and instead, help foundrymen use alternative computational techniques.

Used on a PC, this system employs a cellular automaton technique, which uses a regular 3-D grid of cubic elements identical to that found in a structured finite difference lattice. Each element can be identified numerically regarding its state (solid or liquid) and temperature.

During a solidification simulation, the temperatures of individual elements change progressively as a function of the six nearest neighbor elements according to a deterministic rule. The procedure is akin to a time-stepping, dimensionless, finite difference simulation technique. The appearance and location of macroshrinkage defects are determined simultaneously as solidification proceeds.

Although the rapid cellular automaton technique can determine likely macrofreezing patterns, cutoff and macroshrinkage defect locations (because the model does not attempt to solve the physical heat transfer equations), no predictions are possible regarding actual temperatures or times.

An advantage of this approach is the user is not required to input any physical properties data. Boundary conditions can be varied in the model to qualitatively account for the possibility of different types of materials being in contact with the liquid. In this way, the likely influences of chills/insulators/molding materials on freezing patterns can be evaluated.

Rapid Simulation

The first task is to generate a 3-D model of the casting/feeder/runner geometry. This is done with a menu-driven, user-friendly solid modeler where 3-D shapes can be easily constructed out of cubic elements. Figure 1 shows four views of the completed solid model. This model was created as an example in 10 minutes.

Once the model has been completed, it can then be submitted to the solidification simulator. The simulator progressively predicts the macrofreezing of the casting and the likely macroshrinkage defects or the last regions to freeze. The volume percent freezing contraction or the volume percent of the last regions to freeze is preset by the user with simple menu-driven displays.

Two results can be viewed by the user with a selection of post-processing modules. Figure 2 shows predicted defect locations for a particular casting in a 2-D format. Figure 3 displays computed results for the same shape indicating the pattern of macrofreezing for this geometry.

Figures 4a and 4b compare a cellular automaton output results file with the results of a full numerical simulation for the same geometry, respectively. The numerical method result was obtained using an implicit finite difference program incorporating thermophysical data for casting/mold, etc. It can be seen that the form is consistent with the full numerical model, although it was obtained in less than one-tenth of the time and without needing any thermophysical data.

This example represents an aluminum alloy sand casting and assumes that initially the metal and mold temperatures are uniform (i.e., filling was not simulated).

Once the simulation is run, it is easy to alter the 3-D model of the casting/feeder design and run further simulations. In this way, a variety of potential designs can be evaluated on the computer before castings are produced.

One U.K. iron foundry found that within six months of operating the software, it was achieving a 95% "right-the-first-time" success rate on new castings. Furthermore, it achieved considerable financial savings on already existing castings by redesigning the feeding systems to improve casting yields. In the case of one casting, the foundry increased yield from 62%-72.5% and saved |pounds~12,000 a year.

Benefits of Casting Design

The ability to computer design castings can significantly enhance the image of a foundry in the eyes of potential customers. It also can serve as an educational tool to improve understanding of the casting process for foundry personnel. Foundries and their customers will increasingly expect to see computer-aided design systems used for castings.

Casting design benefits include:

* the ability to design castings on-screen to reduce costs normally incurred with the trial-and-error approach;

* enhancing the "right-the-first-time" design capability of foundries;

* the cost savings realized by reducing design and prototype time and costs, and the ensuing engineering and tooling benefits--time and labor savings that quickly outweigh the initial software capital investment;

* the improved likelihood of obtaining casting design concessions or alterations from the customer. Customers are far more willing to agree to design modifications when presented with computer simulation results.

A First Step?

Foundries taking such an alternative simulation route have incorporated solidification simulation into their methoding procedures in little time and seen definite cost benefits. These findings support the advantages of rapid simulation models for short simulation times, ease of use by in-house engineers, no input of properties data required and low-cost hardware.

Although finite difference/finite element models offer the potential for more powerful prediction of solidification behavior and structural evolution, many areas relating to the physics of solidification are still not well understood or quantifiable. Furthermore, considerable experience needs to be acquired in the operation of these numerical programs, which can, therefore, be expensive to operate.

Consequently, many people are beginning to appreciate the advantages of possibly using rapid simulation methods as the preferred first step in casting design even if they already possess a numerical modeling capability.