identifying the origin of sliver defects in rolled …...defects found in rolled aluminum sheet,...

Since compositional differences appear to not be the source of slivers, scalping of rolling surfaces should belimited to removing the exudation layer and obtaining a level rolling surface. To reduce mechanical damage andmitigate edge cracking, a potential source of debris, scalping sides before hot rolling is recommended, as it willremove the harder and less ductile exudation layer.

Identifying the Origin of Sliver Defects in Rolled Aluminum Sheets

Fahad Bokhari, Trey Eads, Hassan Khan, and Matt LeavittFaculty Advisors: Professors Kevin Trumble and Matthew J. M. KraneIndustrial Sponsors: Kirstyn Sudhoff, Sam Warner, and Ryan LassAcknowledgements: Ginny Hammersmith (Logan Aluminum), Shannon Heidrich (Purdue), and Tim Van Meter (Purdue)

Project Background

Experimental Procedure

Process Recommendations

Production Sliver Characterization

MSE 430-440: Materials Processing and Design

Slivers are elongated surfacedefects found in rolled aluminumsheet, seen in Figs. 4A and 5A,and if not removed, result infailure when drawn into a can(Fig. 1). During Direct Chill (DC)casting, ingot-scale segregation,not correctable by heattreatment, occurs due to thesolute rejection into the moltenaluminum and subsequent fluidmotion[1]. Two layers are formed:a strongly solute-rich layer at thesurface, and a moderatelydepleted layer further from thesurface. These layers areremoved on each rolling face ofthe ingot by scalping up to 2.5cm (1 in.), while the side facesare unaltered. Insufficientscalping may leave remnants ofa depleted layer which mayelongate into sliver defectsduring rolling.

This work is sponsored by Logan Aluminum,Russellville, KY.

Macrosegregation AnalysisComposition profileswere measured fromthe surface into thebulk ingot tocharacterize theexudation and depletedlayer depths and tocompare these depthsto Logan’s scalp depth.The solute enrichmentmeasured 1.1 + 0.3mm for 5182, andcontained up to 2.5times the nominal Mgcontent of the alloy’s(4.5wt%). Due to thelower Mg content of3104 (1wt%), nosignificant compositionvariation near thesurface was found. The5182 exudation layerfuck

[1] Haug, Mo, and Thevik. Inter. J. Heat Mass Transfer (1995).[2] Figure Adapted from D. Eskin, “Macrosegregation in direct-chill casting of aluminium alloys”, (2008).

Production slivers in samples provided by Logan Aluminum were classified as either Smeared (Fig. 4A) orRuptured Surface (Fig. 5A) Slivers. No distinct compositional or metallographic differences between thesmeared surface sliver and the sheet suggests the defect is caused by mechanical damage, whereas a distinctboundary divides the ruptured surface sliver from the sheet. Slivers responsible for failure during drawing (Fig. 1)share topographical characteristics such as striations with ruptured surface slivers.

Figure 5: Example ofruptured surface sliver(A) on surface and (B)in cross section. Lackof compositiondifference betweensliver and bulk metalsuggests the defect iscaused by mechanicaldamage.

Figure 4: (A)smeared sliver onsurface. No signof sliver is foundin cross section(B), suggesting itbeing a surfacediscoloration.

Slivers are common elongated surface defects in rolled aluminum sheet that lead to a significant decrease in yield. Thehypothesized causes of slivers are an insufficient removal (scalping) of ingot surface macrosegregation and mechanicaldamage in hot rolling. Examination of the root causes of sliver defects was performed through a scaled replication of theLogan Aluminum rolling process on aluminum alloys 5182 and 3104 and characterization through metallography andcompositional analysis. Lack of compositional variations on the scale of the depletion layer in production slivers led to theconclusion mechanical damage is the more likely culprit.

Figure 1: Crack in canoriginating at andpropagating through sliverduring deep drawing. +

Figure 6: A thick exudation layer in 5182above the red line, while a dendriticstructure is seen below.

Figure 8: Vickers hardness measurements of the 5182 alloy. Hardness ofthe exudation layer is about 10% higher than in the bulk (subsurface).

Liquid

MushyZone

Solid

Water Jets

Water Cooled Mold

Figure 2: Typical DC Caster. Solidification shrinkagepulls the initial solid shell away from the water cooledmold, allowing partial remelting and exudation to occur.[2]

Figure 7: A thinner exudation layer in3104, ~250 µm, is seen above the red line,while a dendritic structure is seen below.

Ingot Casting

Surface Scalping

Hot Rolling

Cold Rolling

was found to have a higher hardness (Fig. 8), due tohigher solute content, leading to a decrease inductility. This results in edge cracks and a potentialsource of debris on the unscalped sides of the ingot.

To isolate potential causes of sliver formation,samples of a DC cast 5182 alloy ingot were cut intothin bars and hot rolled with 15% reductions.Samples were rolled with different scalp depths inorder to evaluate surface layer effects on finishedsheets:

Over-scalped samples were then rolled with thefollowing modifications :

• Scoring: The rolling face was deeply scratched.• Debris: Fine particles or coarse fragments were

placed onto the rolling face during rolling.

A transverse sample was tested to illustrate the effecthot rolling had at different depths. Findings led tofurther rolling experiments with a larger samplehaving an unscalped and scalped side (edge).

• Unscalped: The as-castsurface remains on therolling face of the sample.

• Logan’s scalp: The rollingface was at the scalp depthtypically used by LoganAluminum.

• Over-scalped: The rollingface was at least 10 cm (4in.) below the as castsurface.

• Transverse: The rollingface contains the exudationlayer at one end and bulkmaterial along the rest.

Figure 3: Sample locationand orientation.

Rolling Experiments

Figure 9: Smeared sliver onsample from coarse debris lostduring rolling.

Figure 12: (A) 5182 edge trimfrom Logan Aluminum comparedto (B) the unscalped edge of thetransverse sample.

.A. .B.

Transverse sample defects were constrained to theexudation layer, implying solute enrichment resultsin defect susceptibility. To confirm this effect, alarger sample was rolled with the exudation layer asone side (edge) and a scalped edge as the other.This sample allowed direct comparison of the effectof ingot side scalping. Shown in Fig. 11, the scalpededge remained smooth and did not crack while theunscalped edge was heavily cracked. The damageon the unscalped side was the same as is seen onedge trim of production sheet (Fig. 12).

The control, scored, and fine debris samplesdisplayed no slivering. Coarse debris caused a sliverattached at one end. A light-colored scar, Fig. 9,formed if coarse debris fell off during rolling. Theexudation layer at the head of the transverse samplebegan to crack and separate (Fig. 10).

Figure 10: Separation of rolled exudation layer from bulk sample.

Figure 11: (A) relatively smoothscalped edge (B) extensive crackingon unscalped edge.

.B..A.

As-cast surface

10 c

m

Rolling Direction