Investigating Cast Iron Defects: Four Foundries' Experiences.

Four foundries share their whats, whys and hows of investigating casting defects and the solutions that were provided to eliminate the defects from production.

Casting defects are a hurdle foundries encounter on a daily basis. Whether the defect is seen during cleaning, finishing or final inspection, the goal always is to understand its cause and eliminate it -- particularly before the customer encounters it in its processing of parts.

This article looks at examples of casting defects from four different foundries as presented in a larger panel at the 1999 AFS Casting Congress. By following these foundries' thought processes from defect detection to elimination, various techniques are uncovered for investigating casting defects.

Metallurgical Comparison

Foundry: Goldens' Foundry and Machine Co., Columbus, Georgia.

Castings: Two piston rings, cast in low-carbon equivalent, high-strength class 40 gray iron via green sand casting.

Defect: A difference in polish luster (Fig. 1) between two piston rings was brought to Goldens' attention by its customer--one from an unidentified foundry ("Fukahara") and the other from Goldens'.

Investigation/Solution: Since the Fukahara casting was capable of being polished to a higher luster than the Goldens' casting, they both were subjected to a laboratory metallurgical comparison to determine if alloy content accounted for this difference.

The submitted machined portions of the piston ring were 0.5-in. thick and 2-in. wide--the Fukahara section measured 3-in. long and the Goldens' 12 in.

Segments from both castings were subjected to Brinell hardness testing in accordance with ASTM E10, and results indicated that both castings had similar hardness values, although the Goldens' casting (201 HB) was slightly harder than the Fukahara casting (187 HB).

Random cross-sections were removed from both castings, mounted in bakelite molds, and ground and polished in accordance with standard metallographic procedures per ASTM E3-95. The resulting samples then were examined unetched and etched with a metallurgical microscope of magnifications up to 1000X. The purpose for this examination was to evaluate and compare the microstructures. The unetched structure revealed that both castings had flake graphite identified as ASTM Type A Size Class 4. The etched structure for both castings revealed a matrix that was essentially pearlite. Microporosity was observed in both castings most likely as the result of microshrinkage sometimes associated with low-carbon equivalent iron.

In addition to the microporosity, the Goldens' casting also displayed evidence of intercellular carbides, which, in many instances, were associated with the porosity. The carbides essentially were iron carbides, however, in many instances, alloy carbides from molybdenum (Mo) (0.25%), phosphorus (0.011 %), vanadium (less than 0.005%) and chromium (Cr) (0.19%) also were detected. The presence of these carbides reflects the alloy content that was determined to represent the Goldens' casting but was not present in the Fukahara casting.

The metallographically prepared samples were further examined using a scanning electron microscope (SEM). This examination revealed that the intercellular carbides appeared to have multiple phases. As a consequence, the carbides were further analyzed with energy dispersive X-ray analysis. This technique is capable of identifying the elements present in a microscopic area. The purpose for this analysis was to identify the source of the carbides detected in the Goldens' casting and to determine the composition of the phases present with the void as revealed with a back-scattered electron image. This image revealed a difference in density of the constituents and showed that the constituents, in conjunction with the voids, have multiple phases. The analysis was conducted on the cross-section away from the carbides to establish a base iron composition, resulting in iron, carbon (C) and silicon detection.

Historically, the formation of carbides in cast iron has been minimized with proper inoculation materials and techniques. The iron also must be capable of responding to the inoculation effect. In this instance, the Goldens' casting had a rather low sulfur (S) level of 0.055%, which was near the minimum level. The desired S level for the best response to inoculation for reducing intercellular carbides in gray iron is 0.05-0.09%. It is possible that an increase in S to the higher end of the desired range would improve response to inoculation sufficiently to overcome the formation of the undesired intercellular carbides. However, the most efficient solution to reducing the polishing defects would be to reduce the elements that are segregating, such as the Mo and Cr.

The results of the investigation indicated that a significant difference in alloy content existed between the two castings, accounting for the difference in response to polishing. The Goldens' casting had been alloyed with Mo and Cr whereas the Fukahara casting had not. These alloying elements segregated to intercellular regions during solidification and were the last to freeze, causing microscopic hard spots. Microporosity and occasional phosphides were associated with the presence of these carbides. These hard carbides, in conjunction with the porosity, created the conditions that resulted in the less than desirable luster on the polished surface of the Goldens' casting.

Since the intercellular carbides and phosphides were the last to freeze, the fact that they remained liquid longer than the rest of the casting also is significant. In this case, these liquid phases also served to provide feed metal to the surrounding metal during solidification. The low carbon equivalent (3.67%) of the Goldens' casting combined with the low S (0.055%) and the relatively high aluminum (0.027%) to create ideal conditions for intercellular segregation and microshrinkage. The consequence of this event was that in addition to the microscopic hard spots, microshrinkage also occurred at these intercellular regions, further detracting from the desired polished luster.

As a consequence of this investigation, it was recommended that Goldens' Foundry consider manufacturing these castings without the alloy additions. It also was recommended that the S content for the Goldens' castings be increased to a range of 0.07-0.09% so that better response to inoculation could be achieved. With improved response to inoculation, the occurrence of undesired carbides can be reduced. Goldens' Foundry reported that the changes they made in reaction to these recommendations resulted in an acceptable finish on the castings that were subsequently supplied to its customer.

Dimensional Evaluation

Foundry: Brillion Iron Works, Inc., Brillion, Wisconsin.

Casting: An 80-55-06 grade ductile iron gear blank (Fig. 2) of a camshaft, produced in green sand, via high-production, vertically parted molding.

Defect: Brillion's customer informed the foundry of the gear blank's dimensional variation (0.537 in.) exceeding its tolerance (0.456-0.486 in.) and its non-uniformity within all mold cavities.

Investigation/Solution: This defect was resulting in a 9% reject rate and an added expense for the customer. In response, the foundry devised a team to address the defect and come up with a solution. This team consisted of Brillion's engineer, superintendent, inspection supervisor, maintenance supervisor, molding unit supervisor, pattern shop supervisor, plant engineer, quality assurance engineer, quality assurance manager and sand system superintendent.

The team's first order of action was to quarantine all inventory and attribute gage 100% of the parts. It also established a coordinate measuring machine program for evaluation of the dimensional data (variable).

The key to the team's defect investigation, though, was its scientific problem-solving, in which members compared the dimensions of the current part to those found at layout for production approval. They also evaluated a mold cluster for dimensional variation and identified the mold cavities of defect concentration.

In its first attempt to identify the problem, the team reviewed process sheet parameters, key process parameter variation in sand and molding, and the foundry's maintenance log for adjustments. It also verified pattern conditions and dimensions. It was determined that the problem did not reside in these areas.

Next, the team took investigative action by conducting several tests, which revealed there was no correlation between sand parameter variation and dimensional results and a direct correlation between the mold hardness and part thickness. There was no change in benchmark dimensional variation with a 0.010in. shim behind the center of the pattern plate, and moving the pouring cup 2.5 in. from the center dispersed the defect concentration to various cavities. The team also discovered a capable dimension on all cavities when moving the pouring cup 4.5 in. from the mold center. After comparing the dimensional capability of gray and ductile iron, the team also found that ductile iron castings are thicker because of greater ferrostatic pressure.

In its root cause analysis, the team determined that the pouring cup "shadow" caused uneven sand filling of the mold, blocking the flow of sand and keeping it from packing up tightly. In addition, a negative correlation existed between mold pitch (thickness) and dimensional size, and a high-squeeze pressure distorted the molds and did not improve their density.

Finally, as part of its Discrepancy Recurrence Prevention, the team standardized the mold pitch, squeeze pressure and sand properties, located the pouring cup to an optimum location, and applied preventive action principles to similar part numbers.

Through its defect analysis, Brillion was able to cut its overall part scrap performance down to 0.2% with only one piece being rejected for dimensional causes.

Macroviews, Photomicrographs and SEMs

Foundry: Grede-St. Cloud, St. Cloud, Minnesota.

Casting: A 65-45-12 grade ductile iron automotive casting for an engine belt drive produced in green sand via high-production, vertically parted flaskless molding. The casting is produced with 12 on a pattern with two mold cavities per riser.

Defect: After being machined by an outside facility, four different defects appeared in this casting--cold shut, gas holes, sand inclusions and stress cracks.

Investigation/Solution: In terms of investigating the defects, Grede-St. Cloud took the "seeing is believing" approach. By using macroviews of the casting from 3-10X magnification, photomicrographs from 25-100X magnification and producing images with an SEM, the foundry was able to isolate the images for analysis (Fig. 3).

The cold shut defect was seen in SEM images that isolated the iron beads that had formed. A cold shut results from an interrupted pour, with the initial metal splashing and cooling (forming iron beads) before the rest of the metal follows and solidifies around it. To eliminate the iron beads and cold shut, the answer was to increase the pouring speed of the molding line's automatic pouring system.

For the gas holes defect, photomicrographs were used. Taken at magnifications of 50 and l00X, these pictures illustrated the relationship between the gas bubble, graphite and lustrous carbon. The gas bubbles form next to the sand mold and the C builds up inside the bubble. As the casting solidifies, the C solidifies onto the iron. Grede's solution to its gas inclusions was to increase the venting on the mold so the gas escapes out of the mold instead of remaining trapped.

The most obvious defect in this casting was the sand grain inclusions that occurred. Viewable with simple macroviews, a stereoscope with 25X magnification clearly shows the embedded grains. The sand inclusion problem can be attributed to a faulty gating system design because as the molten iron flowed through the gating, it eroded the sand, which became trapped as the iron solidified. To detect where the problem in the gating system was, the foundry pulled a casting tree before cutoff to observe where the metal filled an eroded area.

The last defect was a crack, which showed up during machining. The sharp corner on the casting is a stress raiser that cracked on machining. The crack is visible on the macroviews at 3 and 10X. The solution to the stress was to soften the tooling at the corner.

Solidification Modeling

Foundry: Citation Corp., Birmingham, Alabama.

Casting: This high-production ductile iron 8-lb slip yoke is cast in green sand on a cope and drag line with six on per pattern. Feeding is performed by four risers, two of which feed two castings each with the other two feeding a single casting.

Defect: The slip yoke experienced shrink at the contact with the riser. The risers feeding a single casting experienced a higher incidence rate of shrink than the riser feeding two castings.

Investigation/Solution: There were several factors that illustrated the shrink occurring in the casting at the riser contacts. Sometimes open shrink at the riser contact was readily visible. In other instances, the shrink was hidden behind the contact and only visible on X-ray (three castings per production run were X-rayed). Another sign of shrink at the contacts was dispersed porosity. Last, after grinding the riser contact, burn marks (overheated metal) sometimes would appear indicating a void under the surface of the casting.

Once the foundry determined the shrink defect, the first step in the investigation was to determine if the riser was feeding effectively. If the riser wasn't feeding, the shrink defect could have been the result of: a high carbon equivalent, the ingates being too thick, the riser being too big, having too many risers, or too low of a metal temperature in the casting. The foundry investigated the risers at shakeout to see if the riser tops were piping down, and it was determined that the riser was feeding properly. This left three other possibilities for a shrink defect with a riser that does feed--the riser diameter is too small, the carbon equivalent is too low or the temperature of the metal in the casting is too high causing the piping to extend to the contact.

Through further investigation, the carbon equivalent and the metal temperature in the casting were determined to be adequate because the piping didn't extend to the contact. This left as a solution the riser diameter being too small. This reason would account for the increased occurrence of shrink in the single cavity casting with a single riser feeding it.

The riser feeding two different castings had 16 lb of metal flowing through it to feed the two mold cavities, as compared to 8 lb with the riser feeding a single casting. As a result, the riser feeding two castings heated the surrounding sand more, creating more radiant heat in the surrounding sand that allowed the riser to sustain a higher temperature, stay molten and feed the castings for a greater length of time.

The foundry verified the problems with the riser diameter via solidification modeling. This modeling supported the foundry's belief, therefore the foundry began to test various riser diameter sizes. The end result was a larger riser (a 20% increase in diameter). The solidification models in Fig. 4 illustrate the before and after of the riser feeding two cavities.