How to Screw Up a Tensile Test


Roy Lobenhofer


The tensile test is the standard criteria for the acceptance of the iron in most specifications for gray and ductile iron castings. It has been for years and the procedures/standards for conducting the tensile tests have remained virtually unchanged for decades.[i] One would think with that consistency it would be indicative of a test virtually impossible to screw up. But if one would think that, they would be wrong. Almost every foundryman who has received the news of a failed tensile test immediately wondered if the test was screwed up. (The exception is the foundryman who expected it to fail in the first place.) In most cases the test wasn’t screwed up. It was merely weak iron. However, it might be comforting for foundrymen to run through the ways tensile tests can be goofed up so that they can make sure it didn’t happen to them.


Since laboratories never do anything wrong (just ask them), it isn’t possible for me to list the ways tests are screwed up in an order of likelihood. Instead, I’ll merely list them in areas which the errors can occur.




As with any test there is always a possibility of simple clerical errors goofing up the results.




There is always the possibility of grabbing the wrong sample in the foundry and either taking or sending it to the lab. If a commercial lab is used, a number of samples are frequently sent at one time. It’s not hard to imagine someone cross identifying two bars. Similarly, if a commercial lab receives a number of test bars in one shipment, the possibility of cross identifying or mis-identifying a specimen exists. (Actually, a commercial lab’s method for preventing such occurrences is something that should be reviewed when auditing the lab before using their services.)


This type of error is frequently thought of when results are reported that correspond to another grade of iron produced in the foundry. I am confident it has happened in some of the failures I have dealt with. After saying that, I must point out it is very hard to prove or disprove. I have frequently thought a chemical analysis of the bar compared to what was believed to be chemical analysis at the time would settle the question. The problem is chemical analyses of as-cast specimens is costly and determining the correlation between those results and the spectrometer results associated with a specific test bar would require rather extensive testing.


Simple Misreads


As with any human endeavor, it is possible to simply read something incorrectly, Micrometers are frequently used to measure the diameter of test bars, elongations are determined from physical measurements, and tensile machines have read outs indicating the load being applied to the specimen. Those readings can be seen/written down incorrectly. The result is incorrect reports of tensile properties.


While this seems logical, I cannot recall ever seeing an example of a strength result where I was convinced this happened. I theorize errors of this type would typically be so large the technician would catch it before reporting or, conversely, so small I wouldn’t expect a problem.


That isn’t the case with elongation measurements. I have seen cases where   calculations were messed up due to changing gauge lengths and that resulted in incorrect reporting of elongations. In addition, elongation is typically determined by “fitting” the two broken halves together and measuring the stretching that has occurred. There is a lot of discretion used in determining what “fitted” together means. Smashing the halves together and touching them together can make a difference.


In the Lab


Obviously, if you are going to screw up a test, one of the most logical places to do it is in the lab where the test is conducted. While how the tests are to be conducted are spelled out rather explicitly in the specifications, there are opportunities for variability in the testing to occur.




My first contact with tensile testing was as an undergraduate earning some extra money assisting a graduate student work on his thesis that surface finish affected the results of steel tensile tests. I never did see his paper but ever since then when the machining of test bars is mentioned my interest is peaked.


It seems logical to me that better surface finish would result in better properties. On the simplest level if one looks at the surface as a series of peaks and valleys, the diameter will be measured by a micrometer on the peaks, but the effective diameter is really the smaller diameter measured from valley to valley.


The difference between the peaks and valleys is not the most suspect cause of tensile property inaccuracies from machining. Gouges and sharp turns in machining tend to the get that honor. There is something called the “notch effect” created by the gouges and sharp turns. The notch effect, according to the theory, concentrates stresses in the area and, thereby, causes premature failure.


I think I can remember seeing a failed tensile test bar with a gouge, but it certainly hasn’t been a common occurrence. More common has been seeing test bars breaking outside the gauge length making the test invalid. When breakage occurs where the cross section is greater than within the gauge length and the fracture doesn’t reveal a cause, the machining is usually suspected; however, I can’t say I have ever seen proof of those suspicions.




Tensile tests are supposed to be conducted by pulling a specimen apart at a 90o angle to the axis of the specimen. If the pulling is done at something other than 90o, the strength results are diminished because strengths other than tension come into play. (Torque is an example.) While this seems simple and unlikely to occur in practice, I am aware of at least three types of alignment problems occurring in real life.


The most obvious form of misalignment comes when the top of the tensile machine is not completely parallel to the bottom. The deleterious effect of this has been, I’m told, documented in literature. According to information given to me, it takes only a degree or two misalignment to effect the result. Every tensile machine I have ever seen has been a stout piece of equipment and it is hard for me to imagine them going out of alignment through normal use, but when the tensile machine is verified/certified this can, and most likely, should be checked.


I have never had any direct experience with this phenomenon. Hopefully, it is because the machines I have dealt with have been properly aligned, or, it could be I was simply unaware of the problem.


Another form of misalignment can come from the “grips/heads/holders” of the bars in the testing process. One end of a test bar is placed in a grip attached to the bottom of the tensile machine and the other end in a grip attached to the top. The grips are made as matched pairs and mixing the pairs can lead to misalignment.


I did have personal experience with this problem. I was receiving inconsistent results from the commercial laboratory we were using at the time. Sister bars[ii] did not correlate well at times. In discussing this with the laboratory they knew they were always testing them identically. Comparisons with another commercial lab showed far less inconsistencies and led to dropping the inconsistent lab for normal testing. About a year later the salesman for the dropped lab came in trying to convince us to use their lab again. He confessed they found they were not paying proper attention to matching the grip pairs and that resulted in the variation in their results.


Finally, misalignment can also be caused by machining. If the two ends of the bar are not machined in line, problems will arise.


I must confess this problem had never occurred to me. Quantifying the inconsistencies in tensile properties has long been an interest of mine. While working for the American Foundrymen’s’ Society[iii], I worked with a committee that devised a test to help quantify the inconsistencies. As the results came in, I was surprised to see one of the foundries which I admired for their technical expertise showing very erratic results. After the experiment was concluded, the liaison with that foundry came to me and thanked me for setting up the experiment that allowed him to get a new lathe to machine his test bars. He explained he had been suspect there was a wobble in his lathe and the results from the tests and comparison to the other labs proved to his management his point.


Pulling Rate


The rate of loading can affect tensile tests. Many believe if a ductile iron test bar is loaded at a high rate a higher yield strength can be attained. Conversely, it is believe a slower loading rate benefits elongation. To my knowledge there hasn’t been any formal testing to prove these theories and careful reading of ASTM E8 will show there are ranges for the loading rates that should be used.


This trick was something I learned early in my career. At that stage of my career ductile iron pipe were first being made and we struggled getting the yield strength high enough while passing the elongation requirement. I would use a relatively high loading rate until the yield strength was obtained and then slow the rate way down. Did it work? That’s one of the problems with tensile tests. They are destructive. You only have one chance to run the test. You can’t really prove what would happen if you did something different. Sure, an experiment could be set up with multiple tests and possibly find out, but things like that didn’t happen in the foundries where I worked. I guess the only thing I can say to verify the trick, is that I used the system as long as I worked at the pipe shop. That being said, I would doubt severely whether a commercial lab would ever use this technique.


          Undersized Bars


When running tensile tests on bars created in standard test molds there shouldn’t be any concern about undersized test bars. (An undersized test bar is one that is smaller than the standard bar used in a foundry.) Undersized test bars (actually any bar machined directly from a casting) shouldn’t be expected to give “standard” results. The irons are known to be “section sensitive” (more accurately cooling rate sensitive). Bars machined to smaller sizes and/or machined from non-standard sections will result in unusual results. There is documentation of the variance in tensile results from pouring different size test bars from the same iron.


Quite simply, don’t expect standard results from non-standard size test specimens.


In the Foundry


While we foundrymen would like to think we do things perfectly, obviously there are things done in the foundries that can effect tensile tests. Sometimes, the changes occur in a single test and can be looked at as screwing up the test.


Molding Material


The material the molds in which tensile specimens are poured is supposed to be equivalent to the material used to make the castings the tests are supposed represent. I have never quite understood how the heat transfer of a core material mold is the same as a green sand mold, but supposedly it is; therefore, most foundries use tensile molds made from core material. (The core material molds can be made well in advance and the molds stored until needed thereby saving setup times.) It’s logical that molds with different heat transfer would change the properties found in the tests. The making of test molds in advance would seem to take away variability of the molding material.


I did hear of one instance when the molding material was causing inconsistent tensile results. I have no idea how it was discovered, but it did encourage me that some people were paying attention to meeting the tensile specifications. The report came from a Wisconsin foundry who, like most foundries made their test molds well in advance and stored them for future use. They experienced “erratic” results one winter. (I don’t remember being told how erratic the results were,) I can’t imagine how they ever found what they did, but they found they had stored the test molds on a pallet, or rack, in such a manner that the molds were touching an uninsulated wall. Because of heat conduction, or lack thereof, one side of the mold was much colder than the other. This resulted in bars from the same mold have significantly different cooling and, thereby, different properties.


That is the only time I have heard of the molding material actually effecting tensile properties and I wouldn’t have believed this story if it didn’t come from such a credible source. If erratic tensile results are being experienced I wouldn’t spend too much time looking at the molding material used to make the test bar molds.




We now, finally, get to where most of the problems with tensile tests occur the pouring of the iron into the test mold.




There are two significant aspects of the pouring likely to cause problems. The first of these are inclusions. One of the significant functions of the gating systems used in iron foundries is to allow trapping of slag inclusions outside of the casting. Standard test bar mold design does not have a gating or filter system. In the standard test bars pouring is done by what is commonly called pouring through the riser. In relation to the test specimen it is a very large riser. The principle behind this is with the large riser, the specimen being at the bottom, and inclusions being less dense than the iron the inclusions will rise out of the specimen leaving only clean iron. For the most part this theory works fine. Problems arise when too many inclusions are poured into the mold and/or the iron solidifies before the inclusions can float out (cold iron).


It’s been my experience, inclusions are the most common way of screwing up a tensile test. In most cases the inclusions can be seen in the fracture of the specimen after breaking. If so noted by the lab, retesting is certainly appropriate. Along this line, while it isn’t uncommon for inclusions to cause problems, it has been my experience that commercial labs don’t do a very good job of reporting their existence. I recommend retrieving questionable bars from commercial labs to personally verify whether an inclusion was the cause of the problem.




In some ways it seems almost ridiculous to include metal as a way to screw up a tensile test, After all, the metal is supposed to be what we are testing. However, as we indicated at the start of this, especially when there is a failure to meet specification, foundrymen like to think there is a screw up in the test. In my years of dealing with tensile tests by far the most common reason for failures has been the metal itself. Chemistries not being maintained or changing nucleation are the most common causes of tensile test failures.


So, if you’re confident your iron hasn’t changed and you get a weird result, you now have a list of things to check as possible causes, but once you’ve eliminated them, look at your iron processing again. That’s most likely where the problem lies.



[i] The standard most commonly used for tensile testing is ASTM E8

[ii] Standard specimen molds create two tensile bars. One of which is usually tested and the other, which is referred to as the sister, is retained in case there is a problem with the initial test.

[iii] Now The American Foundry Society