Non-Linear Strain Paths (Stress-Based FLCs)

Conventional Forming Limit Curves (FLCs) gained widespread industrial use since being introduced by Dr. Stuart Keeler in the 1960’s. Applications from feasibility analysis to stamping plant troubleshooting use these principles. The strain hardening exponent (n-value) and thickness are inputs into a shortcut to create the curve placement and shape, but this is applicable to only mild steels, conventional High-Strength Steels, and some Advanced High-Strength Steels. Furthermore, this shortcut is an approximation, coming from a best-fit curve generated from data points gathered over multiple grades.

A typical method used in creating most FLCs includes deforming samples of different widths with a 100 mm (4 inch) diameter hemispherical punch – known as the Nakajima method. An alternate approach uses a flat-bottom cylindrical punch, known as the Marciniak method (Figure 1). Independent of the punch shape used, generating FLCs involves measuring the strains resulting from deforming a blank to a formed shape. The conventional FLC plots major strain on the vertical axis against minor strain on the horizontal axis. This FLC applies only to in-plane stretching in linear strain paths, and assumes that there are no through-thickness stress or strain differences. Assessing bendability or cut edge ductility is not possible with this approach.

Figure 1: Punch Shape Used to Create FLCs Result in Through-Thickness Strain Differences Which Influence the Shape and Placement of The FLC ^S-37

Figure 2 compares the FLCs generated by deforming DP980 with the three punch shapes highlighted in Figure 1. Note the higher strains associated with the 50 mm diameter hemispherical punch compared with the strains generated from the 100 mm diameter hemispherical punch. This punch curvature difference impacts the magnitude of the strains that develop through the thickness of the sheet. On samples deformed with a hemispherical punch, the selected strain measurement technique (circle/square grid analysis or Digital Image Correlation, for example) directly measures strains on the outer top surface only, with the middle and inner surface having progressively lower strains as a function of the R/T ratio. A punch or feature with small R/T leads to high strains on the outermost surface. Strains exceeding the FLC on only this outer surface will not lead to necks on the formed panel. Exceeding the FLC through the entire thickness – from the inner surface to the outer surface – must occur for the sample to show a neck.^T-17

Figure 2: FLCs of the same batch of DP980 Showing Dependence on Punch Shape and Curvature.^{S-37, M-15}

In addition to the through-thickness strain differences from the punch curvature, the metal flow differences resulting from the punch shapes leads to directional changes in the strain path taken by the deforming metal. A channel drawn part with a hat-shaped cross section in which there are no features like embossments is likely to have a linear strain path. Forming every other engineered stamped part geometry involves some degree of a non-linear strain path (NLSP).

The importance of strain path and deformation history comes from the changes in the forming limit that occur once metal deformation starts. The black curve in Figure 3 shows the FLC for an alloy generated in a conventional manner with as-received metal, assuming a linear strain path. The red curve results from testing the same metal that initially stretched to an equal-biaxial plastic pre-strain of 0.07. In this strain path, substantially less deformation can occur before reaching the forming limit. However, the strain path changes if the local part contour is different, and that strain path results in a different amount of subsequent deformation prior to necking. The magnitude and direction of the shift changes based on the strain and the orientation relative to the rolling direction. Citation S-38 highlights these curves and presents more examples of the effects of different strain paths. The important conclusion is that the amount of deformation that a metal is capable of withstanding prior to necking changes throughout the forming process and depends on the local part shape (among other variables), and cannot be discerned by using only the conventional strain based FLC.

Figure 3: Experimental FLCs for a linear strain path (in black) and for a bilinear strain path after 0.07 strain in equal biaxial tension in strain space (in red) ^S-38

Figure 4 shows the strain paths associated with the FLCs presented in Figure 2, with along with a magnified portion of one of the curves. This non-linearity is a characteristic of samples formed with a dome, associated with the sample wrapping around the punch during the initial contact and experiencing a combination of biaxial bending and stretching. Citation M-15 presents a method to correct for strain path effects.

Figure 4: Strain Path for FLCs shown in Figure 2. A) 100 mm diameter flat punch; B) 100 mm diameter hemispherical punch; C) 50 mm diameter hemispherical punch; and D) Magnified portion of one curve from Figure 4B showing the non-linearity of the strain path.^{S-37, M-15}

Accounting for tool contact pressure is critical as well, since pressure through the sheet thickness suppresses the onset of necking. Applying this compensated FLC in simulation or in hands-on analysis parts analysis requires modification for the unique characteristics of each part, with appropriate adjustments for local curvature, contact pressure and deformation history. Citations S-37 and M-15 detail methods to compensate for the effects of strain path, curvature, and tool pressure. Figure 5 shows that after incorporating these corrections, the curves condense to one shape independent of the variables used.

Figure 5: As-generated FLCs compared with FLCs after strain path, curvature, and tool contact pressure corrections.^{S-37, M-15}

In summary, FLCs generated from relatively similar simple tools are sensitive to small differences in R/T ratio, incorporation of tool contact pressure, and deviations from a linear strain path. By comparison, engineered stampings require substantially more complex tool shapes with differing degrees of curvature, tool contact pressure, and strain paths all within one part. These complex part shapes contribute to an even wider variation in the yield surface and hardening mechanisms important for simulation, and impacts predictions of formability, springback, and stress analysis.

A common requirement during tooling buyoff – where all strains need to be below the FLC by at least a certain amount called the safety margin – magnifies these challenges. AHSS grades already have low FLCs relative to their lower strength counterparts, so it is critical that the chosen FLC does not further reduce efficient application of these grades. Minimizing sensitivity to the changes in strain path occurring across a complex part requires using a different approach – a FLC with the axes in stress-space rather than the conventional strain-space.

This discussion has centered on conventional strain-based FLCs, which incorporate an assumption of a linear strain path as a flat sheet deforms to the final shape. Stress-based Forming Limit Curves (sFLC or FLSC) are insensitive to deformation history and can be adjusted to reflect the differences in local tool geometry or contact pressure across the stamping. Forming analysis software readily converts conventional FLCs into stress-based units. Figure 6 converts the two strain paths presented in Figure 3 into stress-space, and shows the two experimental stress FLCs generated with different strain paths are independent of the loading history and essentially overlap.  Citations S-38, S-39, S-40 and S-41  contain information about stress-based FLCs, as well as their generation and usage.

Figure 6: After converting the conventional FLCs in Figure 3 to stress-space, the experimental stress-based FLCs show no significant differences.^S-38

Citation H-20 presents a related method to transition from strain-based to stress-based Forming Limit Curves. The proposed stress-based failure criterion postulates that localized necking occurs when a critical normal stress condition is met. This approach adequately describe the experimental strain-based forming limit data in most evaluated materials, failing only with a 3rd Generation AHSS alloy containing a high percentage of retained austenite. For this grade, the authors speculate that a material model more advanced than the one employed in this study will improve correlation.

Accurate simulation requires accurate and complete inputs, including the full range of metal properties, with correct material flow and hardening models, and an understanding of the conditions that will produce failure. Any shortcuts taken increases the likelihood that simulation will not fully match reality for all materials, part shapes, and production processes. A conventional strain-based FLC assumes no effect of part geometry, tool contact pressure, and deformation history – all of which occur on engineered stampings to differing degrees. Analysts should incorporate stress-based FLCs into their simulation with appropriate adjustments to address local geometry and contact pressure to ensure an accurate representation of the metal’s forming characteristics.

For use in the die shop or stamping plant, a growing number of optical systems have built-in features to map strain measurements on to an sFLC. Use caution when employing this approach since these systems measure only the final net strain, and not the strain history as the panel deforms. Proper application involves capturing metal flow from individual breakdown panels and adjusting the FLC accordingly as the panel gets closer to the home position.

Special thanks to Dr. Thomas Stoughton, Technical Fellow, General Motors Research & Development, for assistance in preparing this information.