Simulation Inputs

Simulation Inputs


Predicting metal flow and failure is the essence of sheet metal forming simulation.  Characterizing the stress-strain response to metal flow requires a detailed understanding of when the sheet metal first starts to permanently deform (known as the yield criteria), how the metal strengthens with deformation (the hardening law), and the failure criteria (for example, the forming limit curve). Complicating matters is that each of these responses changes as three-dimensional metal flow occurs, and are functions of temperature and forming speed. 

The ability to simulate these features reliably and accurately requires mathematical constitutive laws that are appropriate for the material and forming environments encountered. Advanced models typically improve prediction accuracy, at the cost of additional numerical computational time and the cost of experimental testing to determine the material constants. Minimizing these costs requires compromises, with some of these indicated in Table I created based on Citations B-16 and R-28.

Table I: Deviations from reality made to reduce simulation costs. Based on Citations B-16 and R-28.

Table I: Deviations from reality made to reduce simulation costs. Based on Citations B-16 and R-28.


Yield Criteria

The yield criteria (also known as the yield surface or yield loci) defines the conditions representing the transition from elastic to plastic deformation.  Assuming uniform metal properties in all directions allows for the use of isotropic yield functions like von Mises or Tresca. A more realistic approach considers anisotropic metal flow behavior, requiring the use of more complex yield functions like those associated with Hill, Barlat, Banabic, or Vegter.   

No one yield function is best suited to characterize all metals. Some yield functions have many required inputs.  For example, “Barlat 2004-18p” has 18 separate parameters leading to improved modeling accuracy – but only when inserting the correct values. Using generic textbook values is easier, but negates the value of the chosen model.  However, determining these variables typically is costly and time-consuming, and requires the use of specialized test equipment.

Hardening Curve

Metals get stronger as they deform, which leads to the term work hardening. The flow stress at any given amount of plastic strain combines the yield strength and the strengthening from work hardening.  In its simplest form, the stress-strain curve from a uniaxial tensile test shows the work hardening of the chosen sheet metal. This approach ignores many of the realities occurring during forming of engineered parts, including bi-directional deformation.

Among the simpler descriptions of flow stress are those from Hollomon, Swift, and Ludvik.  More complex hardening laws are associated with Voce and Hockett-Sherby. 

The strain path followed by the sheet metal influences the hardening. Approaches taken in the Yoshida-Uemori (YU) and the Homogeneous Anisotropic Hardening (HAH) models extend these hardening laws to account for Bauschinger Effect deformations (the bending-unbending associated with travel over beads, radii, and draw walls).

As with the yield criteria, accuracy improves when accounting for three-dimensional metal flow, temperature, and forming speed, and using experimentally determined input parameters for the metal in question rather than generic textbook values. 

Failure Conditions

Defining the failure conditions is the other significant challenge in metal forming simulation. Conventional Forming Limit Curves describe necking failure under certain forming modes, and are easier to understand and apply than alternatives. Complexity and accuracy increase when accounting for non-linear strain paths using stress-based Forming Limit Curves.  Necking failure is not the only type of failure mode encountered. Conventional FLCs cannot predict fracture on tight radii and cut edges, nor can they account for dimensional issues like springback.  For these, failure criteria definitions which are more mathematically complex are appropriate.

Constitutive Laws and Their Influence

on Forming Simulation Accuracy

Many simulation packages allow for an easy selection of constitutive laws, typically through a drop-down menu listing all the built-in choices. This ease potentially translates into applying inappropriate selections unless the simulation analyst has a fundamental understanding of the options, the inputs, and the data generation procedures.

Some examples:

  • The “Keeler Equation” for the estimation of FLC0 has many decades of evidence in being sufficiently accurate when applied to mild steels and conventional high strength steels. The simple inputs of n-value and thickness make this approach particularly attractive.  However, there is ample evidence that using this approach with most advanced high strength steels cannot yield a satisfactory representation of the Forming Limit Curve.
  • Even in cases where it is appropriate to use the Keeler Equation, a key input is the n-value or the strain hardening exponent. This value is calculated as the slope of the (natural logarithm of the true stress):(natural logarithm of the true strain curve). The strain range over which this calculation is made influences the generated n-value, which in turn impacts the calculated value for FLC0.
  • The strain history as measured by the strain path at each location greatly influences the Forming Limit. However, this concept has not gained widespread understanding and use by simulation analysts.
  • A common method to experimentally determine flow curves combines tensile testing results through uniform elongation with higher strain data obtained from biaxial bulge testing. Figure 1 shows a flow curve obtained in this manner for a bake hardenable steel with 220 MPa minimum yield strength.  Shown in Figure 2 is a comparison of the stress-strain response from multiple hardening laws associated with this data, all generated from the same fitting strain range between yield and tensile strength.  Data diverges after uniform elongation, leading to vastly different predictions. Note that the differences between models change depending on the metal grade and the input data, so it is not possible to say that one hardening law will always be more accurate than others.
Figure 1: Flow curves for a bake hardenable steel generated by combinng tensile testing with bulge testing L-20

Figure 1: Flow curves for a bake hardenable steel generated by combining tensile testing with bulge testing.L-20


Figure 2: The chosen hardening law leads to vastly different predictions of stress-strain responses L-20

Figure 2: The chosen hardening law leads to vastly different predictions of stress-strain responses.L-20


  • Analysts often treat Poisson’s Ratio and the Elastic Modulus as constants.  It is well known that the Bauschinger Effect leads to changes in the Elastic Modulus, and therefore impacts springback.  However, there are also significant effects in both Poisson’s Ratio (Figure 3) and the Elastic Modulus (Figure 4) as a function of orientation relative to the rolling direction. Complicating matters is that this effect changes based on the selected metal grade.  
Figure 3:  Poisson’s Ratio as a Function of Orientation for Several Grades (Drawing Steel, DP 590, DP 780, DP 1180, and MS 1700) D-11

Figure 3:  Poisson’s Ratio as a Function of Orientation for Several Grades (Drawing Steel, DP 590, DP 980, DP 1180, and MS 1700) D-11


Figure 4:  Modulus of Elasticity as a Function of Orientation for Several Grades (Drawing Steel, DP 590, DP 780, DP 1180, and MS 1700) D-11

Figure 4:  Modulus of Elasticity as a Function of Orientation for Several Grades (Drawing Steel, DP 590, DP 980, DP 1180, and MS 1700) D-11


Testing to Determine Inputs for Simulation

Complete material card development requires results from many tests, each attempting to replicate one or more aspects of metal flow and failure. Certain models require data from only some of these tests, and no one model typically is best for all metals and forming conditions.  Tests described below include:

  • Tensile testing [room temperature at slow strain rates to elevated temperature with accelerated strain rates]
  • Biaxial bulge testing
  • Biaxial tensile testing
  • Shear testing
  • V-bending testing
  • Tension-compression testing with cyclic loading
  • Friction

Tensile testing is the easiest and most widely available mechanical property evaluation required to generate useful data for metal forming simulation. However, a tensile test provides a complete characterization of material flow only when the engineered part looks like a dogbone and all deformation resulted from pulling the sample in tension from the ends. That is obviously not realistic. Getting tensile test results in more than just the rolling direction helps, but generating those still involves pulling the sample in tension.  Three-dimensional metal flow occurs, and the stress-strain response of the sheet metal changes accordingly.  

The uniaxial tensile test generates a draw deformation strain state since the edges are free to contract.  A plane strain tensile test requires using a modified sample geometry with an increased width and decreased gauge length, 

Forming all steels involves a thermal component, either resulting from friction and deformation during “room temperature” forming or the intentional addition of heat such as used in press hardening. In either case, modeling the response to temperature requires data from tests occurring at the temperature of interest, at appropriate forming speeds.  Thermo-mechanical simulators like Gleeble™ generate such data.

Conventional tensile testing occurs at deformation rates of 0.001/sec. Most production stamping occurs at 10,000x that amount, or 10/sec. Crash events can be 2 orders of magnitude faster, at about 1000/sec.  The stress-strain response varies by both testing speed and grade. Therefore, accurate simulation models require data from higher-speed tensile testing. Typically, generating high speed tensile data involves drop towers or Split Hopkinson Pressure Bars.

A pure uniaxial stress state exists in a tensile test only until reaching uniform elongation and the beginning of necking.  Extrapolating uniaxial tensile data beyond uniform elongation risks introducing inaccuracies in metal flow simulations. Biaxial bulge testing generates the data for yield curve extrapolation beyond uniform elongation. This stretch-forming process deforms the sheet sample into a dome shape using hydraulic pressure, typically exerted by water-based fluids.  Citation I-12 describes a standard test procedure for biaxial bulge testing.

A Marciniak test used to create Forming Limit Curves generates in-plane biaxial strains.  Whereas FLC generation uses 100 mm diameter samples, larger samples allow for extraction of full-size tensile bars.  Although this approach generates samples containing biaxial strains, the extracted samples are tested uniaxially in the conventional manner.

Biaxial tensile testing allows for the determination of the yield locus and the biaxial anisotropy coefficient, which describes the slope of the yield surface at the equi-biaxial stress state. This test uses cruciform-shaped test pieces with parallel slits cut into each arm. Citation I-13 describes a standard test procedure for biaxial tensile testing.  The biaxial anisotropy coefficient can also be determined using the disk compression testing as described in Citation T-21.

Shear testing characterizes the sheet metal in a shear loading condition. There is no consensus on the specimen type or testing method. However, the chosen testing set-up should avoid necking, buckling, and any influence of friction.

V-bending tests determine the strain to fracture under specific loading conditions. Achieving plane strain or plane stress loading requires use of a test sample with features promoting the targeted strain state. 

Tension-compression testing characterizes the Bauschinger Effect.  Multiple cycles of tension-compression loading captures cyclic hardening behavior and elastic modulus decay, both of which improve the accuracy of springback predictions.  Again, no standard procedure exists. The biggest challenge with this test is preventing buckling from occurring during in-plane compressive loading. Related to this is the need to compensate for the friction caused by the anti-buckling mechanism in the stress-strain curves .

Friction is obviously a key factor in how metal flows.  However, there is no one simple value of friction that applies to all surfaces, lubricants, and tooling profiles. The coefficient of friction not only varies from point to point on each stamping but changes during the forming process. Determining the coefficient of friction experimentally is a function of the testing approach used. The method by which analysts incorporate friction into simulations influences the accuracy and applicability of the results of the generated model.

Studies are underway to reduce the costs and challenges of obtaining much of this data. It may be possible, for example, to use Digital Image Correlation (DIC) during a simple uniaxial tensile testing to quantify r-value at high strains, determine the material hardening behavior along with strain rate sensitivity, assess the degradation of Young’s Modulus during unloading, and use the detection of the onset of local neck to help account for non-linear strain path effects.S-110


Application of Advanced Testing to Failure Predictions

Global formability failures occur when the forming strains exceed the necking forming limit throughout the entire thickness of the sheet. Advanced steels are at risk of local formability failures where the forming strains exceed the fracture forming limit at any portion of the thickness of the sheet.

Fracture forming limit curves plot higher than the conventional necking forming limit curves on a graph showing major strain on the vertical axis and minor strain on the horizontal axis.  In conventional steels the gap between the fracture FLC and necking FLC is relatively large, so the part failure is almost always necking.  The forming strains are not high enough to reach the fracture FLC.

In contrast, AHSS grades are characterized by a smaller gap between the necking FLC and the fracture FLC.  Depending on the forming history, part geometry (tight radii), and blank processing (cut edge quality), forming strains may exceed the fracture FLC at an edge or bend before exceeding the necking FLC through-thickness.  In this scenario, the part will fracture without signs of localized necking.

A multi-year study funded by the American Iron and Steel Institute at the University of Waterloo Forming and Crash Lab describes a methodology used for forming and fracture characterization of advanced high strength steels, the details of which can be found in Citations B-11, W-20, B-12, B-13, R-5, N-13 and G-19.

This collection of studies, as well as work coming out of these studies, show that relatively few tests sufficiently characterize forming and fracture of AHSS grades.  These studies considered two 3rd Gen Steels, one with 980MPa tensile strength and one with 1180MPa tensile.

  • The yield surface as generated with the Barlat YLD2000-2d yield surface (Figure 5) comes from:
    • Conventional tensile testing at 0, 22.5, 45, 67.5, and 90 degrees to the rolling direction, determining the yield strength and the r-value;
    • Disc compression tests according to the procedure in Citation T-21 to determine the biaxial R-value, rb.
Figure 5: Tensile testing and disc compression testing generate the Barlat YLD2000-2d yield surface in two 3rd Generation AHSS Grades B-13

Figure 5: Tensile testing and disc compression testing generate the Barlat YLD2000-2d yield surface in two 3rd Generation AHSS Grades B-13


  • Creating the hardening curve uses a procedure detailed in Citations R-5 and N-13, and involves only conventional tensile and shear testing using the procedure included in Citation P-15.
Figure 6: Test geometries for hardening curve generation. Left image: Tensile; Right image: Shear.  N-13

Figure 6: Test geometries for hardening curve generation. Left image: Tensile; Right image: Shear.N-13


  • Characterizing formability involved generating a Forming Limit Curve using Marciniak data or process-corrected Nakazima data. (See our article on non-linear strain paths) and Citation N-13 for explanation of process corrections].  Either approach resulted in acceptable characterizations.
  • Fracture characterization uses four plane stress tests: shear, conical hole expansion, V-bending, and a biaxial dome test.  The result from these tests calibrate the fracture locus describing the stress states at fracture.


Back To Top



Simulation Overview

Simulation Overview

Evaluating sheet metal formability using computer software has been in common industrial use for more than two decades. The current sheet metal forming programs are part of the transition to virtual manufacturing that includes analysis of casting solidification and rolling at the metal production facility, welding, moulding of sheet/fiber composites, automation, and other manufacturing processes. Computer simulation of sheet metal forming is known by several terms, including computerized forming process development and computerized die tryout. 

Many highly developed software programs closely replicate the physical press shop forming of sheet metal stampings.   These programs have proven to be accurate in predictions of blank movement, strains, thinning within the blank, wrinkles, buckles, and global formability concerns of necking strains and forming severity as defined by conventional forming limit curves.  Accurate prediction of local formability related failures such as cut edge expansion is more challenging due to modeling of the production process influence as opposed to the ideal laboratory edge.  Prediction of springback generally provides helpful results in understanding the trends and effects. The quality of springback predictions vary with the specific stamping geometry, the selected metal grade, the input information, and user experience, as discussed in Simulation Inputs.

Virtual forming-process development is ideally suited to the needs of current and potential users of AHSS grades. A full range of analysis capabilities is available to evaluate AHSS behavior in a new stamping.  These programs allow rapid “what-if” scenarios to explore the impacts of different grades of AHSS, alternative processing, or even design optimization.   A Design of Experiments on actual tooling in a physical press shop is limited to only a few variables and may be subject to noise variables clouding the results. In the virtual press shop, changing variables is done with a stroke of the keyboard, and is far easier to undo than permanent changes to the tooling.

Virtual die tryout has numerous advantages, allowing for assessing the viability of part, process, and die design all before cutting the first hard die.  Addressing problems before costly and time-consuming die construction starts leads to improved quality and a better use of resources.

The type of software used depends on the goals and available information at each stage of the process, shown in Figure 1.  At the beginning of the styling to production cycle, feasibility – whether the stamping can even be manufactured – is the key question. With only the 3D CAD file of the final part and material properties, One-Step or inverse codes can rapidly ascertain strain along section lines, thinning, forming severity, trim line-to-blank, hot spots, blank contour, and other key information.  This approach takes the finished part geometry and unfolds that shape to generate a starting blank, calculating the strain between the two shapes (formed vs. flat). Since it starts with the finished geometry rather than the blank, the process is the inverse of reality. All deformation takes place in a single step, or one step, leading to the description of a one step inverse code.  Although this takes reduced computing time, only some simulation packages allow incorporation of the forming process, tooling geometry, and the changes in metal properties associated with deformation.

Figure 1: Software types change during processing stage.

Figure 1: Software types change during processing stage.

Achieving more accurate results involves incremental simulation, where the virtual forming process attempts to replicate reality. This approach models the tools (punch, die, and blankholder) and process parameters (like the blankholder forces, blank shape, and bead geometry, location and restraining forces). Each increment, or time-step, reflects the sheet metal deformation at a different position of the press stroke. Subsequent increments rely on the output from the prior increments. As the quality of the inputs increase, so does the precision of the results.

During selection of process and die design parameters, software evaluates how each new input affects the strains and blank movement (including wrinkles and splits), and generates a press-loading curve.  The analysis creates a visual record of the blank deformation into the final part through a transparent die. Each frame of the video is equivalent to an incremental hit or breakdown stamping.  Problem areas or defects in the final increment of forming can be traced backwards through the forming stages to the initiation of the problem, allowing problems to be addressed when before they even occur. Some software packages allow analysis of multi-stage forming, such as progressive dies, transfer presses, or tandem presses. This virtual environment also shows the effects of trimming and other offal removal on dimensional precision and springback.

AHSS grades are suited for load bearing or crash-sensitive applications, and forming simulation helps to optimize performance.  Previously, the static and dynamic capabilities of part and assembly designs were analyzed using CAD-generated stamping designs with inputs of initial sheet thickness and as-received yield strength. Often the tests results from real parts did not agree with these early analyses because the effects from forming were not incorporated. State-of-the-art applications now model the forming operation first, allowing for local thinning and work hardening to occur.  That point-to-point sheet thickness and strength levels are mapped to the crash simulation inputs, resulting in crash models nearly identical to physical test outputs. Correcting deficiencies of the virtual parts by tool, process, or even part design occurs before tool construction has even begun.

Many simulation packages can evaluate the performance of AHSS grades in many forming environments.  A simple constitutive equation with a single n-value sufficiently approximated the stress-strain response of older grades.  The n-value of AHSS grades changes with strain, so when simulating AHSS grades, input the full stress-strain curve instead of choosing just one n-value.  However, this capability may not be present in some proprietary industrial and university software. Use caution when using these programs to analyze AHSS stampings.

Today’s AHSS grades are not the commodities of yesteryear, but instead are highly engineered products unique to the production equipment and processing route chosen by the steelmaker.  Although many companies may be capable of meeting the minimum and maximum mechanical properties associated with a specific grade, different suppliers may take up a different portion of the acceptable window. Working with your production steel supplier helps ensure you are using company-specific forming data.  

Also remember that there are multiple products associated with a targeted tensile strength.  For example, not only are there different families of 980 MPa minimum tensile strength steels (like dual phase, TRIP, and Q&P), but within each family there are multiple options.  A grade designated as “DP980” may have enhanced global formability as measured with total elongation, enhanced local formability as measured in a hole expansion test, or have a balance between those properties.  The associated material card for simulation will be different, and use of the incorrect card in development could lead to an under-engineered process when attempting to run with production steel.


Key Points

  • A virtual tryout evaluates the stamping and die design, and can detail areas of severe forming, buckles, excessive blank movement, and other undesirable deformation before starting tool construction.
  • Evaluating alternative “what-if” scenarios in a virtual environment is faster and more efficient than grinding and welding on physical tooling, and mistakes have no permanent impact.
  • Design of Experiments (DOE) studies are challenging to do in the physical press shop. The large number input changes required by the process makes it difficult to attribute changes to the intended variables tested or to some noise factor. Simulation allows for easier and faster evaluations when changing one or more parameters.
  • AHSS grades change properties when deformed under different forming conditions. Forming simulation captures these changes to show how the final product will react.
  • Observing the sheet metal transition from a flat blank to the engineered stamping in a transparent die is a valuable troubleshooting tool provided by virtual forming.
  • Simulation codes accurately capture blank motion, thinning strains within the body of the stamping, and other global formability issues like severity compared with the conventional forming limit curve.  However, prediction of local formability concerns like sheared edge stretchability is lacking due to the challenges of capturing all the effects related to creating that cut edge.
  • The virtual forming codes do not accurately predict springback or the success of springback correction procedure because of the lack of accurate data everywhere in the stamping during the entire forming operation. However, users report reduction in recuts of the die from 12 or more to three or four based on information from the code.

A well-defined material card and accurate descriptions of metal flow throughout the stamping are critical for precise springback predictions, as discussed in the section on Simulation Inputs.  Lacking that, many simulation packages can at least provide guidance on the trends which may help guide springback countermeasures.