AHSS stamping simulation: All things being (not) equal

Software, using the right data, should have no trouble simulating the stamping of AHSS

THE FABRICATOR® AUGUST 2008

August 26, 2008

By:

Given complete and accurate material data, software should have no trouble simulating the stamping of advanced high strength steel.

AHSS Stamping Simulation Software

Figure 1 While AHSS does present some major hurdles for stamping, simulation software has no insurmountable obstacles.

When people hear advanced high-strength steel, they think about entering new territory; about reconsidering traditional ways of designing a stamping operation; and about big formability hurdles, including some drastic springback. And for AHSS, die tryout has become even more relevant because simulation technology hasn't evolved to the point where it can accurately simulate forming these advanced materials.

Right? Well, regarding the last-mentioned, not really.

Simulation and Material Data

While AHSS does present some major hurdles for stamping, requiring robust presses and die designs, no insurmountable obstacles exist for software, even in its current state of development (see Figure 1). Many of today's problems arise with material models. Mild steel has well-accepted, well-documented models that provide yield strength; tensile strength; n value, or work-hardening exponent; and other characteristics. If you plug that data into most forming simulation software, it will produce a characteristic model that determines how a material should stretch and form in a die.

With the higher-strength steel, material modeling can fall apart because the methods most often employed in simulation are derived from very basic material "laws"—equations based on empirical observations to allow for behavior predictions, such as the power law. Problems arise when you try to force-fit real AHSS test data into those predictive behavior models. With the power law, for example, common mechanical properties revealed are values for the yield strength, tensile strength, and the work-hardening exponent, which is effectively the slope of the curve between the yield strength and the tensile strength, at the curve's high point. But these curve-fit models are based on certain assumptions. One fit model assumes that the n value is measured between 10 percent and 20 percent of the strain value. However, some high-strength steels don't even achieve 20 percent total elongation. Therefore, the reported n value cannot be assumed to fit the same within the application of the power law (see Figure 2).

Not only are those data points—n value, yield strength, and tensile strength—incomplete for AHSS modeling, they can be inaccurate as well. For instance, n value changes during forming. A big debate continues among metallurgists about how these n values should be reported. Some say there should be an initial n value, followed by an intermediate and terminal n value.

In this light, the greatest hurdle for simulation remains not the software code, but lack of a reliable material model. Much of the time an engineer is left with a conventional material description: martensitic 1100, for instance. The name by itself doesn't give enough information. On top of this, material-naming conventions haven't evolved to the point where they're standard across the industry. What one mill considers typical dual-phase 900 may be different from another mill's dual-phase 900. Besides some formability idiosyncrasies, different areas of the world may refer to these steels differently. Some may interpret the "900" in the previous example as the tensile strength; others say it refers to yield strength.

To complicate the matter further, commercial issues arise between the OEM, the stamper, and the steelmakers. An automaker may require a steelmaker to supply an AHSS with minimum yield strength of 900 MPa. The steel supplier finds a batch of steel that meets that specification. Unfortunately, that batch happens to have a yield strength of 1,200 MPa. Technically, that metal meets the minimum—surpasses it, in fact. But the forming characteristics between 900 and 1,200 obviously change; and without reliable test data, and a reasonable expectation that the material will not so severely exceed a desired window of mechanical properties, simulation cannot capture that difference.

Beyond complications behind the stress-strain behavior (see Figure 3), the next level of complexity comes from variation of these properties resulting from different rolling directions of the coil. Material stretched in the rolling direction—the direction in which the sheet was formed at the mill—will have different forming characteristics than a material stretched across or diagonal to the rolling direction. Say a steel has a minimum plastic-strain ratio, or R value, of 1.1. That mean value (or R bar) takes the average grain characteristic when the material is stretched in three dimensions. It's an average, not an actual characteristic. Metal would have to be perfectly homogenous and isotropic to have such a consistent R value throughout, meaning the material grain would be uniform in every direction—something that never happens. Normally, these values change depending on the rolling direction.

Actual testing would reveal different stress and strain characteristics resulting from the rolling direction. Different rolling directions account for the observed differences in how well a part might stretch when forming across the grain compared to forming along the grain, and this in turn affects strain values. So when the strain varies, the stress level within the part at the end of forming is different. If the stress level at the end of forming is different, then the part springs back differently.

As for how the material might split, the Keeler-Goodwin forming limit curve (FLC) poses problems. The model was established when mild steel was the norm and high strength was in the realm of 340 MPa. Now materials are three to even five times stronger than those tested when the FLC was established. So far there is no universal FLC model that can accept only a few basic AHSS material parameters and predict when the material will fail. Only through rigorous empirical testing can a failure limit be recognized for that batch, coil, heat, or grade of steel—a luxury not always available during the design phase.

Power Law curve diagram

Figure 2 Click image to view larger These graphs depict the power law curve-fit for n-value approximation.

Why AHSS?

A fundamental question many product designers and engineers overlook is, Why choose AHSS? The automotive industry works with it to resist undesirable deformation. Metal, however, doesn't know the difference between desirable and undesirable deformation. Manufacturing steel products requires deformation from the mill all the way through the stamping plant. So when a steel mill rolls a thick, hot plate from several inches down to several millimeters thick, the metal is deforming and, at the same time, resisting that deformation. When the mill further rolls that steel cold to the desired blank thickness, it resists that deformation, further increasing potential for variation. That resistance, together with roll adjustments and other elements in steelmaking, lends itself to material coming into the stamping plant that may vary greatly from expectations.

Here's another fundamental question: Is there really more variation in AHSS than mild steel, or is it just that the variation matters more? It's the latter. If a 200-MPa material has varied strength of ±10 percent, this creates a variability window of only 40 MPa. But the same tolerance in 1,000 MPa creates a variability of a not-so-insignificant 200 MPa.

A Different Kind of Modeling

Modeling AHSS requires complete data. It can't rely just on an n value, a yield strength, and a tensile strength. Today simulation software platforms allow you to import a file that details an entire material model—that is, the entire stress-strain curve—based on hard test data. Material is sent to a tensile test lab where it can undergo a battery of tests. Then the lab sends back raw data, which can be directly used for the material law modeling, eliminating the curve-fit required to extract the n value as well. This allows software to base a simulation not on mere data points, but an entire curve, so it can account for the stretching and compression characteristics of the real material.

The shortcoming of applying this technology to AHSS is not a quantum leap in simulation technology, but the availability and reliability of material models that recognize the nuances in these newer steels. Engineers must have a material model that is more in tune with the actual material characteristics. The simulation technology is here today; it just needs the data.

Not a Linear Problem

One of the biggest challenges with AHSS is springback. What stampers need to know is how accurately the simulation will reflect real-world results. The traditional approach uses a static set of input data, perhaps median yield and median tensile strength. But is that really going to produce the median springback result? Not according to how things work in the real world, where the term "all other things being equal" never really applies.

For an analogy, think of a golfer who on a single swing on the practice range slices the ball 20 degrees to the right. After that first swing he plans a compensation for that slice by aiming 20 degrees to the left, and he continues this for the rest of the day. That compensation often doesn't solve the problem, and the ball veers off course—sometimes straight, sometimes to the left, sometimes even farther right. Why? It's because the solution—aiming to the left—works only "with all other things being equal." And that rarely, if ever, happens. So instead, a golfer might take a series of 30 swings at the driving range to perfect his technique (taking the variation out of the problem), to recognize what combination of factors in his swing gives the least variability.

The same applies in metal forming. Stamping is not a linear problem in which a simple numerical link exists between an undesirable result and a single easy-to-adjust fix. Instead, it is a variable process. Under a given set of circumstances, for instance, a lubrication change might bring about the desired process change, while in other cases the same lubrication change may have a negligible effect.

For every stamping operation, process inputs vary within a statistical range, and today simulation can capture that range. These stochastic simulations (seeFigure 4) do not use just a single yield or tensile strength, or a single binder pressure, and so forth. Instead, software can accept minimum yield, maximum yield, and, from a known standard deviation, calculate simulation through statistical probabilities—adjusting the material law data interactively.

So, rather than having a bank of 15 computers running one simulation to finish a single simulation faster, the same battery of computers can run many versions of the simulation, each with different variables and with each value allowed to vary from predetermined medians. The bank of computers display not just a single prediction, but a range of variation for that result, and even allow you to isolate the predicted median output behavior. This all happens before any cutting tool starts chipping away on steel to machine a die.

Armed with this data, the designer can engineer different variables into the system—varying blank size, binder pressure, bead geometry, tool geometry, and so on—and assess how they affect the outcome. If a specific area of a part is especially critical, a die designer can highlight that area—such as a draw area or edge geometry—and find which process variations are most influential, be it yield strength, blank location, or some other numerical parameter. Through tweaking, the designer can uncover which variables give the most repeatability and focus on controlling them as much as possible.

Stress Stain curve diagram

Figure 3 Click image to view larger As this stress-stain curve illustrates, forming at higher stresses increases springback, esb.

Simulations do consider certain variations, such as material quality or grain structure, which cannot be changed—factors called production noise. But the simulations do uncover the payoff for changing some aspects of the process design to eliminate a noise factor from occurring in the first place. For instance, say controlling a blank width (Y) dimension has the greatest effect on controlling springback. A typical sheared blank width, however, is the result of coil width; sheet comes from a standard master coil and is fed through a shear cutoff. But what if the operation used a blanking die that kept the blank dimensions consistent? Such a consideration might seem crazy at first: Not only are you investing in a blanking die, but you are also wasting material.

But the simulation gives designers foresight: The consistent blanking die will produce more waste at the beginning of the line, but there will be less waste at the end of the line because fewer parts will hit the scrap pile.

How Industry Got Here

Some may argue that simulation has issues with AHSS. They're right, to a point, but those issues arise not from software inadequacies but from a lack of available data about the material. In fact, many would argue that the dual-phase and other AHSS applications in use today wouldn't have gotten to where they are without simulation software.

Even so, these materials do have variables that present serious challenges. Simulation needs actual material data, and material attributes can change from vendor to vendor. There is a need for a robust solution that is insensitive, if not immune, to the noise in the production environment.

Software has long been vaunted as a way to fully eliminate die tryout, and you probably have heard at least one example of when first-hit success was achieved. But as material strengths rise, there are just too many variables. With fuel prices and safety crash standards on the rise, the automotive industry is broadening the palette of what can be formed in the pressroom, and AHSS surely will play a big role.

Stochastic Simulation Diagram

Figure 4 Click image to view larger Stochastic simulations give a range of results to predict how a process will repeat under real-world conditions. These charts, representing two different process setups, show how a material's varying yield strength affects process repeatability. The circles represent specific simulation results. The circles on the left show the process gives no repeatability; the ones on the right show the process repeats more often. (The vertical axis represents displacement in the normal direction, or the direction perpendicular to the reference surface.)



Eric Kam


AutoForm Engineering USA
500 Kirts Blvd.
Suite 113
Troy, MI 48084

Related Companies

Published In...

The Fabricator®

The FABRICATOR® is North America's leading magazine for the metal forming and fabricating industry. The magazine delivers the news, technical articles, and case histories that enable fabricators to do their jobs more efficiently. The FABRICATOR has served the industry since 1971. Print subscriptions are free to qualified persons in North America involved in metal forming and fabricating.

Preview the Digital Edition

Subscribe to The Fabricator®

Read more from this issue

comments powered by Disqus