Editor’s note: Using traditional plastic moldfilling simulation packages to predict the outcome of a MIM process lacks some key elements, according to Gerald Backer of EKK Inc. He has developed an alternative that more accurately models the MIM process. In this excerpted version of Backer’s technical paper, he explains why MIM simulation must rely on a CFD (computational fluid dynamics) solver for optimum results.

Traditionally, designers have predicted moldfilling for the MIM process by using simulation software from the plastics industry. Codes such as C-Mold and Moldflow, with robust and efficient solvers for highly viscous plastic materials, were a natural choice because the binder in MIM feedstocks is generally a highly viscous organic melt. Metal powder added to the organic binder was assumed to have an effect on only two properties—it increased thermal conductivity and modified viscosity of the feedstock.

However, MIM parts diverge from most plastic injection molded parts in more than just the material properties. Differences can be observed in part designs, process parameters, and in the criteria used to identify defects. In some cases, these distinctions make it clear that using a general-purpose plastics code is a less-than-optimum choice. An improved alternative is the use of a general-purpose CFD solver with modifications specifically tailored to the MIM process.

Part Contrasts
MIM and plastic parts vary significantly in the function for which they are designed. Plastic injection moldings are often large, thin-walled parts such as housings, door panels, and instrument panels. For these applications, load bearing capacity and precision manufacturing tolerances are generally not an overriding concern. By comparison, MIM parts are small, strong, and have extremely precise, highly complex designs. A range of consumer and industrial applications for MIM parts exists, two examples being air bag sensors and microwave modules.

The change in scale between plastic injection moldings and MIM parts is also significant. Simulating large thin-walled parts by the commercial plastics codes is accomplished via a solution algorithm that relies on shell elements—2-D elements for which the variation in the third dimension is assumed to follow an analytical relation. For example, channel or fountain flow between opposing walls of the flow channel is assumed. Provided that the flow channels are in fact thin-walled, this 2.5-D approximation is reasonable.

Using shell elements, however, entails some other assumptions. In the thermal energy balance, for example, end effects and corner effects (heat transfer to the mold through corners and edges) are not adequately addressed.

MIM parts, because of their relatively small size, are better characterized by a 3-D representation. For small MIM parts, edge and corner effects on heat transfer are a significant factor in the overall thermal energy balance. Algorithms developed for 3-D hexahedral, wedge, or tetrahedral elements have no error in heat transfer calculations because heat transfer through edges or corners is explicitly modeled.

Feedstock Variations
Flow patterns of MIM feedstock are also more accurately modeled in a 3-D solver. In commercial plastics codes, the assumption of fully developed channel flow implies that the fluid contacts the mold walls on both the upper and lower surfaces of the flow passage. Consequently, common MIM effects such as jetting are not modeled in these codes. These flow patterns are predicted, however, in a 3-D CFD solver.

A second limitation arises when considering the nature of defects that must be predicted by the code. Two common uses for moldfilling codes are to predict the locations where two liquid fronts join together and create a weldline, or the sites where pockets of air are trapped by the motion of the liquid. Like plastics, the MIM process can also encounter these two common types of defects. In addition, though, it can suffer from feedstock that is not homogeneous.

Nonhomogeneous feedstock refers to the tendency of metal or ceramic powders to migrate within the binder, causing some areas to have an excess of powder while others have a deficiency. When these conditions occur at a critical location in the part, it can lead to differential shrinkage during binder burnout or sintering. This, in turn, can cause the part to crack during downstream manufacturing operations, or can lead to high residual stresses in the finished product.

Better Solution
Based on these and other considerations, a 3-D flow solver (Wrafts) was modified to simulate highly viscous flow and to predict segregation of the metal powder during filling.

To verify that the simulation matched reality, an experiment was conducted to measure particle segregation under a given state of applied shear. These experimental results were replicated by the modified Wrafts solver, thus proving its accuracy. The velocity profile for these experiments varies from 117 rpm at the inner core to 0 at the outer wall. Consequently, the location of highest shear is at the inner core while the lowest shear rates are at the outer wall. This difference in shear rate in turn causes a variation in particle concentration.

The modified solver was then applied to a simplified geometry—a corner bracket with a flange and gate—for which some experimental observations of powder segregation are documented. During the simulation, a mixture of metal powder and plastic binder was injected at 30 cm/sec. The mixture spread out as it contacted the wall opposite the gate (Figure 1), but the flow to the upper part of the mold was reduced when the mixture turned the corner (Figure 2) and the majority of the flow proceeded through the lower section of the cavity.

When simulation results were output as a contour plot of the metal powder concentration (Figure 3), they show an increase in powder concentration on the side of the bracket opposite the gate. In addition, a reduction in the metal powder concentration is predicted at the bracket corner. These changes in metal powder concentration were caused by the shear stresses experienced during the flow of the injection molding mixture into the mold. Particles tended to migrate away from locations where shear stresses were high, and accumulated in locations where shear stresses were relatively low.

In actual molding trials, cracking defects were observed at the bracket corner, where metal particle concentration was predicted to be low. Based on this trial, it is likely that the algorithm could be used to provide some guidelines for modifying the part or gating design to minimize powder segregation. For example, the gate could be relocated to the bracket corner, or the radius of the corner could be increased to reduce shear stresses at that location. Simulations for the modified designs could then be performed to determine their effectiveness in reducing the extent of the segregation, and by association, the extent of cracking during the manufacturing process.