Effects of Properties on Fatigue Corrosion/Fatigue 4

[Carburize]

I have some firmly held views on the most important variables affecting fatigue and corrosion-fatigue performance but would appreciate any comments others may have on the following:
How significant is toughness on the fatigue, corrosion-fatigue perfomance of a material?

 

[metengr]

Carburize;
In evaluating fatigue performance of metal, I will stay focused on fatigue crack initiation and propagation (reference to Region I and Region II for fatigue crack behavior) because toughness plays a significant role in the tolerance of fatigue cracks in steel structures (Region III). Corrosion obviously increases the rate of fatigue crack propagation in steel, so in my opinion, this variable would be independent of toughness.

Regarding fatigue crack initiation (Region I), toughness is normally inversely proportional to ultimate tensile strength (UTS) in steel. Since UTS is proportional to fatigue strength (using the approximation that the fatigue strength is approximately 50% of the ultimate tensile strength for steels), as the UTS increases in metals, the fatigue strength also increases proportionally. Thus, toughness plays a significant role in the initiation of fatigue cracks.

For actual fatigue crack propagation in metals, this is really insensitive to microstructure and flow properties. If you evaluate the Paris equation exponents, they are more dictated by material, not by heat treatment.

Thus, for fatigue crack initiation in metals, toughness plays a significant role. Once fatigue cracks have initiated in metals, their growth is dictated by a striation mechanism that is insensitive to microstructure and flow properties.

 

[Carburize]

So you would suggest that as ultimate tensile strength increases ( usually accompanied by a decrease in toughness) the fatigue initiation behavior improves.

 

[metengr]

yes.

 

[Carburize]

I agree.
I think some people attribute too much influence to toughness in fatigue in entirely the wrong direction. In the final 20% of life - propagation and final failure yes -but in the initiation phase I believe that stress range and geometry are the dominant features.

 

[unclesyd]

You need to a add another very big factor the enviroment, all types, chemical, thermal, and others.

 

[Carburize]

unclesyd - I agree - I should have prefaced my last remark with "all other things being equal"

 

[lewtam]

A star for Carburize for starting such a great thread and a star for Metengr for a succinct and definitive answer.

In the past, I have been guilty of placing too much emphasis on retarding crack propagation without going a step further and actually addressing crack initiation. I have since found the method of production - reduction ratio, material cleanliness etc has a profound affect on fatigue properties (again all else equal).

This flies in the face of some industry practice (in my experience) of just 'temper it back a bit' to increase toughness and therefore 'fix' your fatigue problems.

 

[Maui]

Carburize, under plane strain conditions a component will fail catastrophically if the applied stress intensity K exceeds the fracture toughness KIC of the material. For failure to occur by fast fracture the combination of crack length and applied stress must satisfy the equation

K > KIC

such that

S > (E*Gc/3.14*a)^0.5 or S*(3.14*a)^0.5 > (E*Gc)^0.5

Where S is the applied stress, Gc is the critical strain energy release rate, a is the crack length, and E is the elastic modulus. Note that there is also a geometry factor that would normally be included, but since a geometry has not been specified it's value has been set equal to 1 for simplicity. For materials that contain small defects and have high fracture toughness values, the applied stress must be relatively large to induce fast fracture. In these same materials failure may occur at much lower stress levels if they are subjected to cyclic stresses. For readers not familiar with this topic, this type of failure occurs due to a process known as fatigue. Repeated or cyclic stresses cause small defects that are present in the material to grow incrementally until the critical flaw size is reached, and then catastrophic failure occurs.

The fatigue life is the amount of time that a component can be safely used in service before the critical flaw size is reached. Fracture mechanics can be used to predict the fatigue life if the specimen geometry, initial crack length, fracture toughness, and the operating stresses are all known. The rate of crack propagation during fatigue can be expressed as a function of the stress intensity factor as,

da/dN = C*(DKI)^m

where da/dN is the crack growth rate, DKI is the applied stress intensity range, and C and m are constants that depend only upon the material. DKI can be determined in a straightforward manner. The maximum stress intensity KImax is reached at the maximum stress Smax during the fatigue cycle and is given by,

KImax = Smax(3.14*a)^1/2

and at the minimum stress

KImin = Smin(3.14*a)^1/2

It is assumed that Smax and Smin are fixed and do not change during the fatigue cycle. The total range of stress intensity that the material experiences during a complete fatigue cycle is given by,

DKI = KImax - KImin

DKI = (Smax - Smin)*(3.14*a)^1/2

Note that DKI is not a constant, but increases as the crack length increases in response to the applied stresses. Substituting this expression into the initial equation our fatigue crack growth law becomes,

da/dN = C[(Smax - Smin)*(3.14*a)^1/2]^m

There are three regions that generally appear on a typical curve of crack growth rate versus applied stress intensity range. I would show this type of curve here, but that capabilty is not straightforward on this website. At relatively low values of DKI the crack growth rate increases rapidly with a small increase of DKI, and this area is known as region I. In this region, a limiting value of DKI is reached below which no measurable crack growth occurs. This limiting value DKTH is called the threshold stress intensity range. If the combination of initial defect size and applied stress produce a stress intensity range below DKTH then no crack growth will occur and the component should not fail by fatigue. In region II at intermediate values of DKI, the fatigue crack growth law that was derived above is usually obeyed. In region III where DKI approaches DKIC, da/dN increases rapidly with increasing values of DKI and the fatigue crack growth law no longer applies.

Maui

 

[Carburize]

Maui - I agree - at least for the last 20% of the life of a part but lets take a practical case of say a suspension component on an SUV. One item comes in with a Charpy of 20 ft lbs and another with 2 ft lbs. Both parts have the same external geometry, surface finish, etc etc - would you expect a significantly longer fatigue life,in reality, from one part compared to the other,if subjected to the same load spectrum?

 

[Maui]

Carburize, the relationship that was derived above does not apply to the last 20% of the life of a part. It applies to stage II where the vast majority of the useful life of the component is spent. There is a clear and experimentally verifiable relationship between the fracture toughness and fatigue behavior for specific classes of metallic alloys, but there is no clear and reproducible relationship between the Charpy impact toughness and the fatigue life that I am aware of. Charpy testing is based upon shock or impact loading, but fatigue testing is almost universally based upon continusously varying applied stress levels, not impact.

Maui

 

[metengr]

Maui;
Perhaps a bit of clarification for me;
your statement regarding fatigue life in stage II would really apply to low cycle fatigue (LCF) crack propagation in region II.

For high cycle fatigue, the majority of fatigue life is spent in region I (crack initiation) and not in region II (crack propagation).

 

[Maui]

Metengr, what I referred to as region II is the area on a graph of natural log da/dN versus natural log DKI where the fatigue crack growth law described by

da/dN = C*(DKI)^m

is obeyed. In region II on this type of graph, which I would very much like to display here, the behavior would be linear. This can be seen from

ln(da/dN) = ln(C) + m*[ln(DKI)]

The line would have a slope equal to the exponent m. Both high and low cycle fatigue can be described using this approach.

The definitions that we apply for regions I, II, and III may not be the same, which is what I suspect is leading to some confusion. In the definition that I am using for stage I, it is assumed that the threshold value of the stress intensity range DKI has been reached or exceeded so that a crack has already been initiated. The crack is therefore growing in each of these regions as the number of cycles increases. If the definition that you are using for region I includes the area below which the threshold value of DKI is reached and you are dealing with a high cycle fatigue problem then yes, a major portion of the time in service is spent initiating the crack, if it ever forms at all. Using that definition, the amount of time spent in region I can be much greater than the time spent in either region II or III. Using the definition that I have described, the majority of the useful life of the part is usually spent in region II.

Maui

 

[Carburize]

maui - I disagree, the majority of the life of most parts is spent in the initiation phase before the crack growth regime starts.

 

[Maui]

Carburize, unless a crack has already been initiated, the equations that are presented above for crack growth rate (da/dN) and applied stress intensity range (DKI) result in values of zero. This result is consistent with the definitions that I have described for regions I, II, and III. You can't expect to describe mathematically how the crack length increases with respect to the number of applied stress cycles unless a crack is already there to begin with. As I stated in my previous post,

If the definition that you are using for region I includes the area below which the threshold value of DKI is reached and you are dealing with a high cycle fatigue problem then yes, a major portion of the time in service is spent initiating the crack, if it ever forms at all. Using that definition, the amount of time spent in region I can be much greater than the time spent in either region II or III.

If the applied stress intensity range is below the threshold value for crack initiation, then a crack will not be initiated and the component in question will not fail by fatigue. If the threshold value for the applied stress intensity range is exceeded then a crack will be initiated. Depending upon the values of the variables involved (number of loading cycles, min and max applied stress, component geometry, material, etc.), the amount of time required to initiate the crack may be much greater than the remaining life of the part after the crack has been started. However, the opposite may also occur - the crack may be initiated in the very first stress cycle, and the majority of the remaining life of the part is spent in propagating the crack. I have personally seen examples of each of these behaviors.

If the part in question is designed properly, then the potential for fatigue failure will be anticipated and the designer will incorporate the necessary features into his design to prevent the initiation of a fatigue crack. Features such as reducing the presence of stress concentrations, fine surface finish, etc. will help to increase the life of the component. Making a blanket statement regarding how components fail by fatigue such as

the majority of the life of most parts is spent in the initiation phase before the crack growth regime starts

is something that I would never attempt to do.

 

[redpicker]

I really am enjoying this thread. It _is_ a controversial subject. I believe most ofthe controversy results because at first glance it seems that the scientific analysis produces conclusions that contradict pratical observations.

Maui, I don't believe that anyone disagrees with your analysis. Most students of fatigue behavior have been exposed to this type of analysis and alot of them even understand it. I think a logical fallacy occurs when applying this research to pratical applications, however.

The fracture mechanics approach, the way I understand it, says that crack growth rates are directly related to the critical stress intensity (Kc), and that Kc is related to KIc (where I am making a distinction between the loading evnironment at the crack tip -Kc- and the mode I loading used in KIc determinations). Furthermore, since CVN values have been shown to be related to KIc values, the conclusion is that materials with higher CVN values will have slower crack growth rates and, therefore, longer fatigue lifes.

It is this conclusion that, I believe, lead to the begining of this thread. That is, pratical experience does not support this conclusion. Components made from materials with higher CVN values do not demonstrate any longer life in fatigue or corrosion fatigue situations (aside from the obvious situation of a material with a toughness so low that a small fatigue crack propagates rapidly to brittle failure).

I think the basic reason for this apparent parodox is Carburize's statement

the majority of the life of most parts is spent in the initiation phase before the crack growth regime starts

Not only is this rather easily made, but it is supported by industry practice. In general, comoponents that are routinely inspected for fatigue cracks are removed from service if a crack of any size is detected. Obviously, the majority (nearly all of it) of the component's life is spent in the initiation phase.

To futher (or perhaps sidetrack?) this discussion, I'd like to ask Carburize's originial question, except substutite "Yield Strength" for toughness, That is

How signifigant is yield strength on fatigue, corrosion fatigue strength of a component (keeping other factors equal, such as material chemistry, environment, loading, etc...)? For example, I have two pump shafts. One from 90,000 psi yield material and one from 130,000 yield material, both have > 40 Ft-Lbs LCVN. Which could be expected to give longer life in fatigue/corrosion fatigue applications?

 

[NickE]

redpicker- from my understanding yeild strength is not related to fatigue strength. I have always understood it to be tensile strength. I have been told this is due to the crack not being able to grow if the stresses at the tip are below the UTS. This would make sense since if the stresses are above yield yet below tensile the material surrounding the crack tip will yield causing a stress reduction.

 

[redpicker]

OK, I guess that's my mistake. For my example, I assumed quenched and tempered low alloy steel; figure a yield to tensile ratio of 88%, so the 90,000 psi yield material would have a UTS of 102,000 psi and the 130,000 psi yield material would have a UTS of 148,000 psi. would changing the UTS by 45% affect the fatigue life?

 

[metengr]

Yes, it would. It is a well established fact that the fatigue limit (which is the stress associated with the horizontal portion of the S-N curve) of certain steels is directly related to ultimate tensile strength as mentioned in the above post and also see page 432 of "Deformation and Fracture Mechanics of Engineering Materials", by Hertzberg.

The desire would be to use higher strength materials in fatigue applications. The problem is when you increase the tensile strength, the toughness deceases and you have increased susceptibility to other fracture mechanisms like brittle fracture (from reduced toughness) or an increase in susceptibility to failure from environmental affects (stress corrosion cracking). Thus, you need to optimize the toughness and tensile strength.

So, to answer your question in simple terms, with ALL things being equal, the pump shaft with the higher UTS will have increased resistance to fatigue.

 

[redpicker]

I know that it is "a well established fact that the fatigue limit (which is the stress associated with the horizontal portion of the S-N curve) of certain steels is directly related to ultimate tensile strength as mentioned in the above post", which is the reason for my question; not the answer. The development of the S-N curves are typically performed on carefully machined (and polished) specimens that are tested in dry air at constant temperature under completely reversed bending (no axial loading). In real-life, however, actual components are not used in such ideal conditions. Axial loading exists. Mill surfaces are used. Corrosive environments are always present. When these conditions are introduced to the standard RR Moore test, the existence of a "fatigue limit" seems to disappear.

I work with oilfield tubulars and fatigue is a constant concern. Full-size testing is very expensive, time consuming, and difficult to control important variables. For these reasons, most decisions are made from data produced from the laboratory test data.

I have seen test results of full size materials indicate that the 1,000,000 cycle fatigue life limit for steels with the above strengths reported as 20,000 psi regardless of tensile strength. Because of the expense (and the likelyhood that the results, like I mentioned, will defy "conventional wisdom") additional testing to verify or contradict these results is not likely to occur anytime soon.

When encountering a field problem due to fatigue, it is nearly always suggested that increasing the strength of the material will reduce the problem. I don't think this is effective. It seems that the only real solution involves reducing the alternating stresses.