Figuring Out Why Tensile Tests Fail

Everyone who has been responsible for making iron for more than a couple of days has faced the dilemma of trying to figure out why the lab reported a tensile test result that did not meet the specified requirements. Too many times the responsible person has looked at the results, “knew” everything was done right, decided the test wasn’t done right, threw the results out and had another bar tested.

The tensile strength requirements are part of virtually every casting order. If a foundry is serious about giving their customers what they order, each failure will be investigated to determine if poor results are truly representative of that iron, and if they are, what needs to be done to prevent it happening again.

I’ve developed a series of questions to ask when faced with a tensile that can lead to what should be done.

The first is should you be surprised about the failure? Let’s face it, if your average tensile strength for a class 40 is 40,750 psi, you shouldn’t be surprised if you get failures. If the standard deviation of previous tests is 750 psi[i], statistics tell us to expect about 20% of the tests to fail.

In order to answer this question, it is necessary to know the average and standard deviation of the tensile strength for each grade of iron produced. It’s been my experience that foundries who have taken the time to know that information have also made the effort to have their processes in control enough that getting a failure is a surprise.

If the statistics say it’s not surprising the failure occurred, it means there is a systemic problem and what is “normally done” needs to be examined and improved. That is what this collective effort has been directed towards. I’d recommend reading all of the sections and redesigning the metal melting and processing accordingly.

If the failure is a surprise, there are more questions that must be asked and answered starting with Does the hardness support the tensile test results? Hardness and tensile strength are related when things are behaving normally. Typically, the higher the hardness the stronger the iron will be. (In ductile iron, the higher the hardness the lower the elongation will be as well.)

If the hardness does not support the tensile strength reported, I’d recommend examining the specimen that was tested. Look for flaws on the fracture surface, tooling marks on the machined surface of the bar and if the fracture occurred in the prescribed area.

Flaws in the fracture are typically slag, but whatever they are, the flaws will not have the strength of the iron and thereby effectively reduce the area of the test bar. The flaws make the test inaccurate, so a retest should be performed, and the data from the initial test ignored.

Tool marks potentially provide dual causes for reducing reported tensile strength. The most severe problem caused by some tool marks is that they can create a notch effect. A notch can cause reduced test results by creating a place for the fracture to start.

The tool marks reduce reported tensile strength even without creating a notch effect. Tool marks reduce readings by giving a false reading of the diameter. When using almost any measuring tool and especially the micrometers used to measure the diameter of tensile specimens, the measurement obtained is of the maximum diameter. Unfortunately fractures typically occur on the minimum diameter. Normally this effect is negligible, but it becomes more significant as the tool marks become deeper and as the diameter of the test specimen becomes smaller.

If significant tool marks are found when examining a failed sample, it is entirely reasonable to throw out the results. Of course, a test of the sister[ii] is in order.

The final examination should be to determine the position of the fracture on the tensile specimen. If it did not occur within the middle third, it is not a valid test and a retest should be performed.

If no visible cause for the failure can be seen, the next question to ask is are there any chemical abnormalities? As pointed out in other sections of this endeavor, the chemical composition of the iron does have a great impact on the properties of iron. There is always the possibility when the amount of any element falls outside of the normal ranges experienced by a foundry, it can be the cause of tensile problems. Therefore, each element normally measured should be compared to the averages and standard deviations developed from successful tests.

We also know nucleation is important to tensile strength in the irons. Unfortunately, measuring nucleation is not very easy. Inferences need to be drawn in order to determine if a change in nucleation was the cause of the problem. Typically, nucleation is thought of as the inoculant that is added to iron immediately prior to pouring. The easy answer frequently given for tensile problems is that the inoculant was not added. To answer the question “was the inoculant added?" I check the silicon and aluminum analysis. Typical inoculants are silicon-based and use calcium and aluminum to produce the inoculating effect. While calcium is not typically measured, silicon and aluminum are. If those two elements are in normal ranges, I am left with the assumption that the inoculation was added properly.

A new consideration, at least for me, has come along in the last few years regarding nucleation. It's called hereditary nucleation. It refers to the nuclei present in the iron prior to the addition of the inoculant. Once again, we are faced with the difficulty of not having a physical measurement of this. We are left to inferring if this could possibly be a source of a tensile problem. In the early stages of such thought, our efforts concentrated on the charge materials. From the "good old days" foundrymen knew that pig iron contain more nuclei than purchased cast scrap, and cast scrap contained more nuclei then steel. Therefore, we looked at the charge makeup and asked if there was anything unusual in the amounts of the charge components.

Even more recently, we started looking at the loss of this hereditary nucleation over time as a possible cause of tensile problems. A number of years ago a problem with iron held in channel furnaces over the weekend was noted. The first gray iron poured on Monday morning was found to be much weaker than expected although the inoculation was done properly and the chemistries were within the normal range. It led to the term "Monday morning iron." What has been found recently is that it does not take a weekend to develop a problem.

The loss of nucleation in liquid iron is a time and temperature phenomenon. The longer the iron is held the more nucleation loss will occur. Also the higher the temperature at which the iron is held the faster the nucleation will be lost. I have noted a decrease in tensile properties in gray iron starting with holding times less than an hour. Therefore, another question we must ask is was the iron held longer than is typical?

The final questions that we ask regarding tensile problems is does the microstructure support the poor results. While many feel that this should be the first question asked, I have found that most foundries do not do microstructures frequently enough to determine whether a specific sample is unusual or not. In order to effectively use microstructures, a database of good samples needs to be developed so that when a failure is examined, differences from the norm can be detected.

I have an additional problem with putting microstructures higher on my analysis list. If we determine that the microstructure is abnormal, we are left with the problem of determining what caused the microstructure.

If all of these questions are answered, there's a good likelihood the cause of the problem will be identified. Unfortunately, that's not always the case. I would say in far too high percentage of the analyses I have conducted, I've come away not knowing a cause. (If I were forced to make a guess at the percentage of not knowing a cause after analysis, I currently put it in the 25% range.) In later years, with new technologies and information, I've looked back on previously unknown causes and said, "I bet that's what caused it."

My frustration with this type of analysis goes beyond not coming up with an answer. It comes from watching nothing being done when an answer is found. If a cause is found, it is because something in the process allowed that cause. If the process is not changed, it can be expected to be performing the analysis again with the same cause being found in the future.

I have developed an Excel spreadsheet to assist in evaluating the cause of tensile failures. I would be happy to email it to anyone if asked. Email me at rwl@lobenhofer.com.

[i] It should be noted a standard deviation of 750 psi is virtually impossible. It is far more feasible that a normal foundry would have a standard deviation 3 to 5 times that large. See the part of this titled “What is Good Control?”.

[ii] In gray and ductile iron the standard test specimens produce two samples. Typically, one sample is used for testing and the other is retained in case of problems with the testing. The bar not tested is referred to as the “sister” in many foundries.