Chapter 12: Fatigue of Metals

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement not replaced the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

Welcome to the Deep Dive.

Today, we are taking a deep, concentrated plunge into one of the most critical concepts in materials, science, and mechanical engineering, the fatigue of metals.

And this isn't about that sudden, you know,

spectacular failure from one massive load.

This is something else entirely.

It is.

It's the slow, silent killer of engineering components failure that comes from repeated stress cycling.

So that's our mission today to really unpack that.

That's the whole game.

Our mission is to provide an intensive, structured summary of this core mechanical metallurgy chapter.

We're going to, you know, walk through the definitions, the math, the curves, all of it, all of it, making sure you can see how these equations translate into safe, real world component of design.

It's really tailored for that engineering student context.

Okay.

So let's start with the sheer scope of this problem.

This isn't some niche, obscure failure mode, is it?

Oh, far from it.

Fatigue failure has actually been recognized for a long time since the railways in the 1830s,

but its significance today is just enormous.

Mechanical fatigue accounts for at least 90 % of all service failures that are due to mechanical causes.

90%.

That's almost everything.

It's almost everything.

Think about that.

Nine out of 10 times a rotating shaft or a pressurized pipe fails because of mechanical stress.

It's fatigue.

And it's insidious, right?

It doesn't give you a warning.

That's the word for it.

It's insidious because it happens without any gross deformation.

The material looks perfectly fine right up until the moment of catastrophe.

No warning at all.

And the visual evidence, the fracture surface itself, that tells a crucial story for any failure analysis.

It tells the whole story.

If you look at a fatigue fracture surface, and this is something engineers learn to spot instantly, you usually see two very distinct regions.

Two regions that map out the entire life history of the part.

Precisely.

The fracture surface usually lies perpendicular or normal to the direction of the principal tensile stress.

And that first region is smooth, almost polished or rubbed looking.

This corresponds to the slow progressive growth of the crack.

And within that smooth region, that's where you see the famous.

The beach marks.

You often find a series of concentric rings or what we call beach marks.

Oh, the beach marks.

They're like the components diary, aren't they?

They absolutely are.

A perfect analogy.

Those beach marks or rings, they mark the successive positions of the crack front.

They show you exactly where the crack was after a certain amount of time or a certain number of load applications.

So if a machine was shut down over the weekend, you might actually see a more pronounced line.

You might.

They're often visible macroscopically.

If the loading sequence was irregular, like those shut down and startup periods, they are the absolute smoking gun for fatigue.

Okay.

So that's the slow growth region and the second region.

The second area is rough.

It's dull.

It looks fractured.

This is the sudden final fracture area.

So that's the moment at all.

Let go.

That's the moment.

It happens when the crack has grown so large that the remaining cross section of the component just can't bear the load anymore.

Even the maximum static load.

And that's when you get that final rapid ductile failure.

Exactly.

And critically, we see that failure almost always initiates at some form of stress Right.

Like a sharp corner or a knot?

A sharp corner, a fillet, a keyway, or even a microscopic inclusion deep inside the metal itself.

Anywhere stress can build up.

So before we dive into the numbers, let's just nail down the essential chemistry for this silent killer to strike.

What three things must be present for a fatigue failure?

The sources define three basic factors that are absolutely necessary.

You can't have fatigue without them.

Okay.

What's number one?

First, you must have a maximum tensile stress component that is of a sufficiently high value.

Tensile stress is what pulls the crack open.

Makes sense.

Second, you need a large enough variation or fluctuation in the applied stress.

The cycling action itself.

A steady load won't do it.

And third.

Third, you have to have a sufficiently large number of cycles of that applied stress.

Lots and lots of repetitions.

So if you eliminate any one of those three, you're safe.

You eliminate fatigue failure.

But if you meet all three, failure is, you know, pretty much inevitable given enough time and enough reversals.

Okay.

So now we can transition into defining the language of fatigue mathematically.

Because these stress cycles, they aren't all created equal.

Not at all.

And we need precise parameters to characterize the fluctuating conditions we're subjecting our component to.

We typically categorize them based on their mean value and amplitude.

Okay.

So what's the first type?

The first and most idealized is what we call reverse stress.

Think of a simple sinusoidal wave where the tension and compression are perfectly balanced.

Symmetrical.

Exactly.

This is often what we try to achieve in the lab with the classic RR Moore rotating beam fatigue machine.

Mathematically, the key feature here is that the mean stress sigma is zero.

But in the real world, that's pretty rare.

So you have repeated stress.

Right.

Which is far more common.

In repeated stress, the maximum stress sigma max and the minimum stress sigma min are unequal.

So they don't balance around zero.

They don't.

They might both be positive.

So tension, tension cycling, they could both be negative or they might span zero.

But the critical point is that the average stress sigma is not zero.

And then there's a really messy one.

The messiest category, irregular or random stress.

This is reality for complex parts.

Think about a bridge.

It's getting wind gusts, traffic temperature changes or an aircraft wing hitting turbulence.

Unpredictable.

Completely.

That random loading requires some pretty advanced analysis to break it down into something we can work with.

All right.

Let's get to the mathematical bedrock now.

We need five essential stress parameters to describe any repeating cycle.

And they all come from just two values, right?

Sigma max and sigma min.

That's all you need to start.

We'll start with a range of stress sigma.

Okay.

Sigma.

What is it physically?

It's the total algebraic difference between the maximum and minimum stress.

So sigma equals sigma max minus sigma min.

It simply defines the total distance or the total swing in stress magnitude that the material sees during one cycle.

Total swing.

Got it.

Next up, probably the most important one, alternating stress or stress amplitude, sigma.

This is the big one.

Sigma is defined as one half of the stress range.

So sigma equals sigma max minus sigma min, all divided by two.

And physically, this is?

This represents the variable part of the load.

It's the stress that's actually driving the fatigue crack forward.

When we look at fatigue limits later, we are almost always solving for the maximum allowable value of this, of sigma.

And the counterpoint to the alternating stress is the mean stress, sigma.

Right.

Sigma m is the algebraic mean of the max and mean stress.

Sigma max plus sigma min divided by two.

So this is the steady component.

It's the steady component, the baseline that the alternating stress is fluctuating around.

If sigma m is positive, the component is under a net tensile load baseline.

And as we'll see later with things like the Goodman diagram, a positive sigma m is a really bad thing for fatigue life.

It's a major reducer of fatigue life.

Hugely important.

Okay, now we get to the ratios.

These are super useful for comparing different tests.

First is the stress ratio, r.

R is just sigma min divided by sigma max.

This simple ratio is like a fingerprint for the test or the application.

How so?

Well, if r equals minus one, it means sigma min is the exact negative of sigma max.

That defines the completely reversed stress cycle, our ideal lab condition where sigma m is zero.

And if r is zero?

If r is zero, the stress goes from zero to a maximum positive value and back to zero.

That's a pure pulsating tension cycle.

So knowing r immediately tells you the type of loading scenario you're dealing with.

And finally, the amplitude ratio, a.

A is sigma a divided by sigma.

This relates the fluctuating component to the steady component.

It can also be expressed directly in terms of r, which is handy.

It's a equals one minus r over one plus r.

And this is useful and it's highly useful in low cycle fatigue analysis.

But the main point is that understanding the interplay between alternating stress and mean stress quantified by these ratios is foundational.

It lets us move beyond the simple lab tests where r is minus one to designing parts that operate under complex non -zero mean stress conditions.

Which is basically every real world application.

From bridge trusses to engine pistons, yes.

OK, we've established the stress state.

Now let's talk about the fundamental graphical tool for fatigue analysis.

The SN curve,

the stress versus cycles to failure curve.

This is the bedrock of high cycle fatigue design.

It absolutely is.

This is the basic plot and it's derived from 10, sometimes hundreds, of individual fatigue tests run at different stress levels.

OK, so describe the axis for us.

On the vertical axis, we plot stress, which we call s.

And it's typically the alternating stress sigma.

And on the horizontal axis?

We plot the cycles to failure n.

And we always, always use a logarithmic scale for n.

Why is that?

Because fatigue life spans so many orders of magnitude.

It can go from,

say, a thousand cycles all the way up to 10 million or even more.

A linear scale would just be unusable.

I think an important note here is that the values plotted on a basic SN curve are usually nominal stresses.

A critical point, yes.

When you look at these raw SN plots, you have to remember they represent an ideal, polished, perfect lab specimen.

Like a small rotating beam.

Exactly.

With no mean stress, so r equals minus one.

They represent the material's inherent resistance to fatigue.

We use this ideal data as a starting point, and then we have to apply correction factors later to account for the real world notches, size, surface finish, all of that.

And when we look at the curve, we can really see two major regimes based on how long the part lasts.

We divide fatigue behavior into high cycle fatigue, or HCF, and low cycle fatigue, or LCF.

HCF is the long life stuff.

Right.

HCF is typically defined as failure occurring above, say, 10 to the 5 cycles.

In this regime, the stresses are low, generally below the material's yield strength.

So the overall deformation is elastic.

Predominantly elastic, yes.

But the microscopic mechanism that causes the crack is still localized plastic strain.

That's a key detail.

Then you have the more dangerous shorter life regime.

That's LCF, low cycle fatigue.

This is generally when failure happens in fewer than 10 to the 4, 10 to the 5 cycles.

Here, the stresses are much higher, potentially even exceeding the yield strength.

And that leads to gross plastic deformation.

Right across the cross -section.

And because of this massive plastic component, we can't use simple stress control anymore.

This regime demands a total shift in our thinking to the strain life approach, which we'll get into a bit later.

Now, probably the most famous feature of the SN curve, particularly for structural engineering, is the emergence of the fatigue limit, SE, sometimes called the endurance limit.

Yes.

For ferrous metals, so carbon steel, many alloy steels.

The SN curve exhibits this characteristic flattening out.

It goes down and then it goes horizontal.

A plateau.

A plateau.

The stress level corresponding to this plateau is the fatigue limit.

The idea is that the material has infinite life if the applied stress is kept below this limit.

So the goal for any steel component that's meant to last for decades, like a bridge beam or a machine shaft.

Is to ensure the operating stress always remains safely below this value.

That's the whole basis of the infinite life design philosophy.

For steel, this plateau usually shows up reliably after about a million cycles.

But this beautiful concept of infinite life doesn't hold for everything, does it?

It does not.

This is a critical distinction.

For non -ferrous alloys, so think aluminum, copper, magnesium alloys, the SN curve does not flatten out.

It just keeps sloping downward, even if it's very gradually, out to a hundred million cycles and beyond.

So if I'm designing an aluminum aircraft wing, I can never truly guarantee it will last forever, even at a really low stress.

That's the critical difference.

You can't.

For non -ferrous materials, we don't define a true fatigue limit.

Instead, we define a fatigue strength for a specified large number of cycles, typically 10 million or 100 million.

So designing with aluminum is inherently a finite life problem.

It is.

And this distinction completely changes your design philosophy, your maintenance, your inspection schedules, especially in an industry like aerospace.

Okay.

We have to pause here and discuss a massive practical challenge in all of this variability.

The SN curve we just described is just an average.

As engineers, we have to recognize that fatigue life and the fatigue limit are statistical quantities.

That's one of the hardest lessons for a new engineer.

It's inherently probabilistic.

You can take a hundred specimens, cut from the exact same bar of steel, test them at the exact same stress level, and the cycles to failure will vary significantly.

For a huge scatter.

A huge scatter.

And this happens because of all these microscopic factors, tiny surface imperfections, unavoidable non -metallic inclusions, slight variations in grain structure, small residual stresses from manufacturing.

And these all act as potential crack initiation sites.

They're all little stress risers waiting to start a crack.

So how do we possibly account for that variability in a real design?

To manage this uncertainty, fatigue data has to be presented statistically.

We often use a plot that shows curves of constant probability of failure.

So instead of one average SM curve, you plot a family of curves.

Yeah.

Imagine the standard stress versus log N plot.

You'll have a curve for P equals 0 .50.

That's your average life, but you'll also have one for P equals 0 .01, which means only 1 % of specimens fail before that point.

That's high reliability.

And another for P equals 0 .99.

So if I'm a designer and I pick a design stress based on that P equals 0 .50 curve, half my components are going to fail prematurely.

Exactly.

Which is not a great outcome.

So for critical components, the engineer has to choose a curve corresponding to a very low probability failure, say P equals 0 .001, to ensure a high level of reliability.

And the fatigue limit itself isn't a sharp line either.

It's a distribution.

It's a whole distribution.

For a high strength alloy steel, the range between a 50 % failure probability and a 5 % failure probability can easily span 80 MPa in stress, which means engineers have to use rigorous statistical methods like the staircase method or probit analysis to determine the fatigue limit with acceptable confidence and reliability.

You can't just eyeball it.

Okay.

That covers the scatter under the ideal lab condition.

Now let's bring back the mean stress, sigma -hem.

So far, our data assumes the mean stress is 0, but virtually every real component has a steady tensile or compressive load.

How does that baseline shift affect the alternating stress we can safely apply?

When sigma -hem is non -zero, it fundamentally changes the game.

In what way?

If the mean stress is positive, so it's in tension, the allowable alternating stress, sigma, that the component can endure for a fixed life is significantly reduced.

And why is that?

Because that tensile mean stress is constantly pulling the material apart, making it easier for the alternating stress to propagate a crack.

It's giving the crack a head start.

And conversely, if the mean stress is compressive, if it's compressive or negative, the fatigue limit is actually increased.

The component can sustain a higher alternating load because that background compression helps keep any microscopic cracks squeezed shut.

So this means we need a way to combine these two stresses, the alternating and the mean, into a single failure criterion.

We need a graphical solution for this.

And that's where the Goodman diagram comes in.

This diagram is the classic graphical tool used to figure out the limiting range of stress that can be sustained without failure when a mean stress is present.

How is it set up?

It plots alternating stress, sigma, on the vertical axis against the mean stress, sigma,

on the horizontal axis.

And the standard approach for constructing this diagram is quite conservative, right?

It is.

What you do is you draw a straight line, known as the Goodman line, from the fatigue limit for completely reversed stress, that's sigma e, plotted up on the vertical axis where sigma m is zero.

Right.

And you draw it straight down to the ultimate tensile strength, sigma u, on the mean stress axis.

And that line defines the safe zone.

That line defines the boundary of safe operating conditions.

If the combination of your calculated sigma and sigma falls below this line, you are theoretically safe.

It's a beautifully simple linear and conservative approximation.

But engineering data is rarely perfectly linear.

So if we look at actual test results, the Goodman line is just one of several ways to model this.

It is.

If we plot sigma versus sigma, there are a few other common models.

What's the most conservative one?

That would be the Soderbergh line.

This one draws a line from the fatigue limit, sigma e, not to the ultimate strength, but to the yield strength, sigma nought.

And since yield is always lower than ultimate.

This line is far inside the others.

It provides the highest safety margin, especially against any static yielding.

Okay, so you have Soderbergh, you have the Goodman line.

Then you have the Gerber parabola.

This curve recognizes that the experimental data for ductile metals often fits a parabolic shape better than a straight line.

So it's not linear.

No, it's a curve that between sigma e and sigma u.

It's less conservative than the linear Goodman line, especially in that middle range of mean stresses.

But it's often a better predictor of actual failure for some materials.

So we've got linear parabolic based on yield or ultimate strength.

How do engineers handle all these choices?

We can generalize them.

We can capture them all using one fundamental structure.

The generalized mean stress equation.

Okay, what does that look like?

The equation elegantly captures both the linear and parabolic relationships.

It's sigma a equals sigma e times the quantity one minus in parentheses sigma m over sigma u raised to the power of x.

Let's break that down.

Sigma m is the alternating stress we're trying to find the limit for.

Sigma e is our ideal lab fatigue limit.

Sigma m is our calculated mean stress and sigma u is the ultimate tensile strength.

And that exponent x is the key.

How so?

If you set x equal to one, you get the linear Goodman line.

If you set x equals two, you get the nonlinear Gerber parabola.

And for the super conservative Soderbergh line?

You just substitute the yield strength sigma naught in place of sigma u inside that bracketed term.

This framework gives you a flexible calculated way to set your design limits.

Let's take a conceptual walk through a sample problem using verbal reasoning, say calculating the necessary bar diameter for infinite life under some fluctuating axial load.

Okay, so the engineer first uses static equilibrium to define the nominal stresses.

You calculate your sigma max and sigma min in terms of the unknown bar area a.

Right, so stress is force over area.

Exactly, and that immediately defines your sigma m and sigma both in terms of one over a.

And you pick your failure criterion.

Right, let's say you choose the conservative Goodman line, so x equals one.

You then plug your calculated sigma m, which is in that generalized equation.

And you solve it for the allowable sigma.

You solve for the required allowable sigma, which will be a number.

Now you have a numerical value for the maximum sigma allowed.

You just equate that number to your expression for sigma in terms of one over a, and you solve for a.

And from the area, you get the required diameter.

Exactly.

It's a classic combination problem.

Yeah.

You're solving for geometry using a fatigue criterion as your constraint.

All right, let's shift gears now.

We're moving from the high cycle fatigue regime, which is mostly elastic, to the low cycle fatigue or LCF regime, where gross plastic strain occurs.

This is critical for parts that undergo large constrained temperature fluctuations.

Think about certain components in a jet engine or thermal processing equipment.

And the key conceptual difference here is that LCF analysis is often strain controlled.

That's the key.

When a material is constrained, like a pipe attached firmly at both ends, a temperature change forces a certain strain amplitude.

If that strain is large enough to be plastic, we have to analyze the resulting stress strain cycle, not just the stress amplitude alone.

And we visualize this using the hysteresis loop.

Right, under constant strain cycling.

Imagine cycling the material between two fixed strain limits.

When you first load it in tension, it follows the normal stress strain curve.

But when you unload and reverse the load into the compression, the path doesn't just go back down the same line, it traces a loop.

And this is partly due to the Bauschiner effect.

Exactly.

The material yields earlier in the reverse direction than you'd expect.

It's like the material remembers the previous load history.

Plastic yielding in one direction lowers the subsequent yield strength in the opposite direction.

And that loop itself contains all the critical information we need.

It does.

The vertical height of that loop is your stress range, delta sigma.

The total horizontal width is your total strain range, delta epsilon.

And that total strain is made of two parts.

Right.

It's the sum of the elastic strain component, delta epsilon E, and the plastic strain component, delta epsilon P, which is the width of the loop at zero stress.

Now, because we have all this large plastic deformation happening, the material itself can actually change its properties during the cycling.

It can, which leads to cyclic hardening or softening.

This is a fascinating phenomenon.

The material isn't static.

So what happens?

Depending on its initial condition, a metal might cyclically harden or soften.

If a metal starts out soft, say it's annealed, the stress required to maintain a constant strain amplitude might actually increase with each cycle.

It hardens until it reaches a stable state.

And the opposite can happen, too.

Conversely, if it starts highly hardened, the stress required might decrease.

It softens.

This cycling continues until the hysteresis loop stabilizes, which usually takes about 100 cycles or so.

That stable state determines the long -term behavior of the material.

And once that loop stabilizes, we can define the cyclic stress strain curve.

A vital curve.

It's derived by connecting the stable tips of the hysteresis loops from various strain amplitude tests.

And crucially, the cyclic curve is often substantially different from the simple monotonic single -low tensile stress strain curve.

So you can't just use your standard tensile test data for LCF analysis?

You can't.

It's insufficient.

And this stable cyclic behavior, it can be described by a power curve that looks a bit like the Ramberg -Osgood equation for static loading.

It does.

The power curve is defined by delta sigma equals k prime times delta epsilon p to the n prime power.

Okay, let's define those terms.

K prime is the cyclic strength coefficient, and n prime is the cyclic strain hardening exponent.

K prime tends to be proportional to the yield strength,

and n prime is often a small number, maybe between 0 .10 and 0 .20 for cyclically hardened materials.

These are essential inputs for calculating LCF life.

This leads us directly to the LCF life prediction tool itself, which is dominated by that plastic strain component.

This is the famous Coffin -Manson relation.

The Coffin -Manson relationship governs the plastic damage component.

It's a straight line when you plot it on log log axis, plotting the plastic strain range versus cycles to failure.

And the equation is?

The equation is delta epsilon p over 2 equals epsilon f prime times 2n, all raised to the power of c.

Okay, why do we use 2n?

2n represents the number of strain reversals to failure.

One cycle has two reversals, right?

Tension to compression is one cycle, but it's two reversals.

I see.

And the other terms?

The epsilon f prime term is the fatigue ductility coefficient.

It's the strain intercept at 2n equals 1, and it's often correlated with the true fracture strain.

And c is the fatigue ductility exponent.

And that exponent is usually steep.

Very steep.

Typically ranges from minus 0 .5 to minus 0 .7 for many metals.

The steepness tells you that even a small increase in plastic strain amplitude drastically reduces life.

It really confirms the high damaging effect of plastic deformation.

Okay, so the Coffin -Manson relation covers the plastic component, which dominates in LCF.

But we need a complete model that spans the entire fatigue life, incorporating the elastic behavior that dominates in HCF.

For that, we turn to Baskin's relation, which handles the high cycle elastic dominated regime.

What's that equation?

Baskin's relation connects the elastic strain amplitude to life.

It's sigma a, which is also delta epsilon e times e over 2, equals sigma f prime times 2n to the power of b.

So we're basically linking the elastic stress amplitude to the number of reversals with a power law.

That's all it is.

Here, sigma f prime is the fatigue strength coefficient, which is the stress intercept at 2n equals 1.

And b is the fatigue strength exponent.

And this exponent b is a much shallower number.

Much shallower.

It's a small negative value, typically between minus 0 .05 and minus 0 .12.

And that shallower slope reflects that elastic strain is much, much less damaging than plastic strain.

The real power of the modern strain life approach is how it puts these two laws together.

The total strain life equation is the comprehensive model.

It is.

It's the superposition of the elastic, which is Baskin's part, and the plastic, which is Coffin Manson's part.

So you just add them up.

You just add them up.

The total strain amplitude is the sum.

So delta epsilon over 2 equals the Baskin term plus the Coffin Manson term.

This single equation is valid across the entire spectrum of fatigue life, from one reversal all the way out to infinite life.

It's universally applicable.

Let's visualize this on a log clot of strain amplitude versus reversals to failure.

What would that look like?

If you plot the elastic term alone, you get a shallow straight line with a slope of B.

If you plot the plastic term alone, you get a steep straight line with a slope of C.

And the total strain curve is the sum of those two.

It's the sum of the two.

So at very short lives, where you have high strain amplitude,

the steep plastic line completely dominates the behavior.

And at very long lives.

At very long lives, with low strain amplitude, the shallow elastic line takes over.

And the intersection point of those two lines is critical.

That's the transition life, 2NT.

Yes, 2NT is the reversal life where the plastic strain component is exactly equal to the elastic strain component.

So what does that mean for a designer?

It means that for lives shorter than that transition life, you absolutely must use the full equation, or at least focus heavily on the plastic term.

For lives longer than that, in the high cycle regime,

the plastic term becomes negligible.

And you can usually just use the elastic term, which simplifies the calculation back toward the old stress life approach.

Let's walk through the reasoning for worked example one.

This is where we calculate the LCF life N for a stainless steel component.

We're given the total strain range, delta epsilon, and the material properties.

Okay, the primary challenge in this type of problem is that strain is your control variable, but your life prediction model, Coff and Manson, relies on the plastic strain.

So you have to separate them?

You have to separate the total strain.

So the strategy is always to first use the nominal stress limits to calculate the elastic strain range, delta epsilon E, using Hooke's Law.

Delta epsilon E is just delta sigma over E.

And once you have the elastic part, the plastic part is just subtraction.

Simple subtraction.

Delta epsilon P equals the total delta epsilon minus the elastic delta epsilon E.

Now you have the plastic damage term.

Right, and now you can use the Coff and Manson relation.

Correct.

Once you have the plastic strain amplitude, which is delta epsilon P over 2, you plug that value, along with the known fatigue -dictility coefficient and exponent, directly into the Coff and Manson relation.

You solve for 2N, then divide by 2 to get the cycles to failure N.

So the complexity isn't just in the formula, but in correctly isolating the severely damaging plastic strain component from the total measured strain.

That's the key step.

We move now from the overall curve behavior to the physical process of failure, moving from the macro curves to the microstructure itself.

How does the metal actually break down internally?

The process is typically broken into four stages.

Right.

The first stage is crack initiation.

This is the formation of the very first microscopic flaw, usually at a free surface or an inclusion.

And then it starts to grow.

Yes, that's the second stage, stage I crack growth.

This is where the crack develops along crystallographic planes of high shear stress.

This movement is very slow, often limited to just a few grain diameters.

Stage II is where the crack truly becomes macroscopic and dangerous.

Stage II crack growth is the major propagation stage.

The crack now ignores the crystal structure and just grows in the direction normal to the maximum applied tensile stress.

The path of least resistance?

The path of least resistance.

This stage accounts for the vast majority of the fatigue life in low cycle applications.

And finally, stage III.

Stage III is ultimate ductile failure.

This is when the remaining cross section just can't hold the load anymore and it yields rapidly and catastrophically.

And it's important to stress that the relative time spent in these stages varies wildly depending on the type of fatigue.

Oh, absolutely.

In HCF, long life, crack initiation and stage I growth can consume 90 % or more of the total life.

So most of the life is spent just getting the crack started.

Exactly.

But in LCF, short life, initiation is rapid and the bulk of the life is spent in that propagation stage, in stage II.

Okay, so focusing on initiation.

It often begins at the surface via the formation of these microscopic ridges and grooves called slip band extrusions and intrusions.

It's like a microscopic pumping mechanism.

Systematic fine slip movements on the order of just one manometer per cycle are occurring within the grains at the surface.

So what are these extrusions and intrusions?

The extrusions are where materials is lifted out and the intrusions are where material is pushed inward.

This creates an irreversible surface discontinuity that acts as a strong stress riser and that riser quickly evolves into an embryonic fatigue crack.

Once that crack gets into stage II, we see those characteristic fatigue striations on the fracture surface.

These are the visual proof of cyclic failure.

Each striation, each one of those tiny lines, marks the crack growth that occurred from a single application of stress, one cycle.

So the spacing between them tells you how fast the crack was growing.

It directly indicates the local rate of crack growth, dad N, during that part of the component's life.

It's incredible.

And a microscopic mechanism driving this rhythmic advance is known as the layered mechanism.

How does that work?

You have to think of the crack tip not as a static feature, but as a microscopic wedge that is constantly sharpening and blunting.

During the tensile half of the cycle, the crack tip plastically deforms and blunts, pushing the material forward.

Then when the load reverses and becomes compressive, the crack faces close.

But that newly created surface from the tension cycle buckles slightly and pushes the tip forward in a sharpening action.

So it's a constant blunting and sharpening process.

That's what allows the crack to advance incrementally with every single cycle, creating that distinct striation pattern.

This detailed understanding of propagation leads us directly to the modern approach for life prediction,

fracture mechanics and the Paris law.

We're focusing on stage two crack growth and the stress intensity factor range, delta K.

Right.

Delta K is the thermodynamic driving force for crack propagation.

It combines the stress range you're applying and the existing crack length, A.

How do you calculate it?

For a simple geometry, the range is defined by delta K equals K max minus K min.

And using a structural factor, Y, the calculation is delta K equals Y times delta sigma times the square root of pi times A.

And there's a critical practical note here about compressive stresses.

A very important one.

If the minimum stress in your cycle, sigma min, is compressive, we assume the crack closes up.

So we take K min as zero because compressive stresses can't drive a crack forward.

They can't pull it open.

Okay.

So if we plot the crack growth rate, that N, against this stress intensity factor range, delta K, we get a characteristic sigmoidal curve on a log log plot.

Right.

And this S -shaped curve divides into three essential regions, each one dictated by a different physical mechanism.

What's the first region?

Region I is the low delta K region.

It's bounded by the threshold delta K delta K.

Below this threshold, the crack growth rate is extremely slow.

For all practical purposes, the cracks are considered non -propagating.

So if you can keep your applied delta K below that threshold, the flaw is dormant.

You're safe.

Then you have region two.

This is the middle region where the log log plot is a straight line.

This is the Paris law region.

And this is governed by that simple power relationship.

The famous one.

Dad N equals A times delta K to the power of P A and P are material constants.

And the exponent P is typically three or four for structural steels.

This linear relationship is the key tool for life prediction in the damage tolerant approach.

And finally, region three.

Region three is the high delta K region where the crack growth rate just accelerates rapidly.

Here, the maximum stress intensity factor, K max, is approaching Kc, the fracture toughness of the material.

And that's when it's all over.

The remaining ligament quickly yields, and unstable crack growth leads immediately to failure.

So if we know the material constants A and P and the geometry of the component, how do we use Paris law to calculate the remaining life, NF, from an existing initial crack?

Since the crack growth rate, Dad N, is a function of the crack length, A, we have to integrate the Paris law.

You rearrange the Paris law to separate the variables.

So you have Dn on one side.

And you integrate this from your initial crack size Ai to the final critical crack size Af.

The integrated solution gives you the total number of cycles, NF,

spent propagating the crack.

This integral is the absolute backbone of the damage tolerant design philosophy.

Let's walk through the reasoning for worked example two.

We're calculating the cycles to failure for a mild steel plate with an existing edge crack under constant amplitude loading.

Okay, it's a multi -step approach.

First, you have to determine the final critical crack size Af.

How do you do that?

You calculate this using the material's fracture toughness Kc and the maximum applied stress Sigmo max, accounting for the specific crack geometry.

Kc is the maximum stress intensity factor the material can sustain before it fails rapidly.

So once you have Af, you have your integration limits.

You have your limits.

The initial crack size Ai and the critical size Af.

You then take the material constants A and P, which are found empirically for that steel, and you plug them along with Ai, Af, and the stress range delta sigma into the integrated Paris law formula.

And that gives you the total propagation life in cycles.

It does.

The core insight here is that the damage tolerant approach requires defining the failure point first, which is Af, before you can calculate the path to get there, which is NF.

We noted earlier that fatigue failure almost always initiates at a stress concentration point a notch, a hole, a fillet.

These geometric irregularities significantly reduce fatigue strength.

We need a way to quantify that reduction.

And we must distinguish clearly between two concepts here.

First, there's the theoretical stress concentration factor Kt.

OK.

This is derived purely from linear elasticity and geometry.

It's the theoretical ratio of the maximum stress right at the root of the notch to the nominal stress in the section next to it.

And the second concept.

The second is the fatigue notch factor Kf.

This is the observed reduction in the fatigue limit due to the presence of the notch.

This one is derived experimentally.

And the key thing is that Kf is always less than or equal to Kt.

Always.

So why?

Why doesn't the notch always cause the full theoretical stress increase under cyclic loading?

Why is the material often kinder to us than elasticity theory predicts?

Because of localized plasticity.

Small notches, particularly in ductile materials, may not cause the full theoretical reduction in the fatigue limit because the material could accommodate and redistribute that stress through localized plastic deformation right at the notch root.

And this leads to the definition of notch sensitivity, Q.

Right.

Q quantifies how sensitive a material is to a notch.

It ranges from 0 to 1.

And the formula.

Q equals Kf minus 1, all divided by Kt minus 1.

So if Q is 1.

If Q is 1, the notch has its full theoretical effect.

Kf equals Kt, and the material is highly sensitive.

If Q is 0, the notch has no effect on the fatigue limit.

Kf equals 1, and the material is completely insensitive.

What dictates whether Q is high or low?

What makes a material sensitive?

Primarily, it's the material strength and the notch radius.

We observe that highly heat -treated high -strength materials, which tend to be less ductile, show much higher sensitivity.

Their Q often approaches one.

And softer materials.

Softer, more ductile materials, like certain aluminum alloys, show much lower sensitivity.

Their Q is often less than one.

And it's interesting.

For very, very small notch radii,

the sensitivity factor Q actually decreases again.

Why is that?

Because the region of high stress is so tiny that there are very few flaws that exist in that small volume to initiate a crack.

OK.

So if we don't want to use Q, there's an alternative empirical approach to calculating Kf, known as Neuber's relation.

Yes.

Neuber's relation provides a practical path to the fatigue notch factor.

It's Kf equals 1 plus the quantity Kt minus 1 divided by 1 plus the square root of rho prime over Rr.

Let's define those terms.

Here, R is the contour radius at the notch root.

And rho prime is a material constant that relates to the strength and microstructural size of the metal.

This approach incorporates the material's inherent resistance to localized stress fields directly into the calculation of Kf.

Let's apply this conceptually with worked example three.

This one deals with a shaft that has a hole, and it's under a combined steady axial load and a fluctuating bending moment.

OK.

This scenario combines all the complexity,

a geometric concentration, and mean stress effects.

So step one.

First, we determine the nominal stresses.

Sigma m comes from the steady axial load, and sigma f comes from the fluctuating bending.

Step two is the notch.

Second, we calculate the fatigue notch factor, Kf, using either Q or Nuber's relation, incorporating the Kt for the hole geometry.

Then you apply it.

Third, we adjust the alternating stress by this factor.

The effective fatigue stress is Kf times sigma.

And crucially, we apply Kf to the alternating component only.

Why only the alternating part?

Because the factor Kf only affects the failure initiation that's driven by the cycling.

It doesn't affect the overall static mean stress.

Makes sense.

And finally, we use the Goodman -Lein criterion.

But we compare the nominal mean stress, sigma,

against the reduced allowable alternating stress, which is sigma e divided by Kf.

The generalized equation then helps us solve for the diameter we need to achieve infinite life under that specific combination of loads.

OK.

Beyond mean stress and notches, there are several other practical factors that significantly influence fatigue strength in the design process.

Let's start with the size effect.

This is a big one.

The intuitive rule is that laboratory tests on small polished specimens tend to overpredict the performance of large machine members.

So bigger is not better here.

Why does size matter?

It's primarily because of statistics and stress gradients.

As the size of the component increases, so does the volume of material that's subjected to high stress.

Which increases the chance of a flaw being in the wrong place.

It increases the probability that a critical flaw,

a large inclusion, or a surface defect exists in that highly stressed volume.

Secondly, in bending or torsion, the stress gradient in a larger component is shallower, meaning a larger volume experiences near maximum stress.

And the data shows this effect is significant.

It is.

The fatigue limit can drop by 10 % or more as the diameter increases from just a few millimeters to 50 millimeters.

And the source material provides an equation based on the stressed volume correlation.

Yes.

The equation is sigma T1 equals sigma T0 times the ratio of V over V0, all raised to the power of minus 0 .034.

So this links the fatigue limit of the small test specimen to the actual component.

Exactly.

Sigma T0 and V0 are for the test specimen.

Sigma T1 and V are for the actual component.

Where V is the volume stress to at least 95 % of the max stress.

The larger the stress volume V, the lower the allowable fatigue limit sigma T1.

It's a direct recognition of the statistical nature of flaws in a large volume.

Next, let's look at surface effects.

This is paramount since cracks virtually always start at the surface.

The surface condition is highly, highly sensitive.

We look at two things, roughness and residual stress.

Let's start with roughness.

Smoothly polished specimens give you the best fatigue performance, period.

Rougher surfaces, like those left by machining or hot rolling, introduce micro -naunches that act as initiation sites.

And they significantly reduce the fatigue strength.

They do.

And this reduction is quantified by a surface reduction factor, CF.

For high -strength steels, which are more notch -sensitive, the penalty for a rough surface is much, much greater than it is for a soft material.

And the other major surface factor, which is one of the most powerful engineering tools we have for increasing fatigue life, is residual stress.

Oh, absolutely.

The ability to intentionally introduce favorable compressive residual stress patterns at the surface is huge.

So processes like shot peening or surface rolling.

Those processes force the surface layer into a state of permanent compression.

What's the benefit of that?

How does it help?

Well, since fatigue cracking is driven by tensile stress, these compressive residual stresses, sigma r, are algebraically summed with the applied tensile stresses, sigma max, right at the point where failure wants to initiate.

So it cancels out some of the bad tensile stress.

It produces a significantly lower effective maximum tensile stress right where the crack wants to start.

By offsetting the applied tension with this locked -in compression, you can dramatically improve both the fatigue limit and the surface life.

Okay, let's move to the real world of cumulative fatigue damage.

Most components see variable amplitude loading.

The stress level is constantly changing.

We need a way to predict failure under a history of different loads.

The standard, simplest estimation tool for this is the linear damage rule known as Miner's Rule.

Miner's Rule, what does it propose?

It proposes that failure occurs when the sum of the cycle ratios equals one.

So the summation of ni over ni equals one.

Let's define those terms.

ni is the number of cycles you've experienced at a specific stress level i, and capital ni is the total life if the component were subjected only to that stress level i.

It's simple, but it assumes damage accumulates linearly, which we know isn't always true.

That's its fatal flaw.

Miner's Rule completely ignores the effect of the loading sequence.

Give me an example.

For instance, testing shows that applying a period of understressing, a low load below the fatigue limit early in life, may actually consume less damage than predicted.

Sometimes it even improves the material's fatigue limit due to localized strengthening.

And a big overload can be extra damaging.

Conversely, a few large tensile overloads applied later in life can dramatically reduce the remaining life.

Because these overloads introduce residual tensile stresses near the crack tip,

which then accelerate subsequent crack growth even at lower stress levels.

So for accurate prediction under random loading, you need something more complex.

You have to use much more complex cycle counting methods like the rain flow method.

Okay, we should probably define the rain flow method here, since it's so crucial for analyzing real world random data.

The rain flow method is a complex cycle counting algorithm.

It takes a messy, irregular, real world stress history, like the strain measurements from a machine operating in the field, and it breaks it down into a series of constant amplitude load cycles that are equivalent in damage to the original history.

So it lets you apply a miner's rule to complex data.

It creates a damage equivalent history that you can then analyze.

Finally, let's touch on some of the key metallurgical variables.

Fatigue properties are very strongly correlated with tensile properties.

For cast iron and steel, the fatigue limit is roughly 50 % of the ultimate tensile strength.

A good rule of thumb.

A good rule of thumb.

For non -ferrous metals, it's closer to 35%.

And you can see this clearly when you plot the endurance limit against Rockwell C hardness.

As hardness increases, so does the fatigue limit up to a certain point.

What about microstructure?

A fine grain size generally promotes higher fatigue life, particularly in HDF because cracks have more boundaries they have to traverse.

It slows them down.

But inclusions are the enemy.

The presence of non -metallic inclusions is the enemy of fatigue life.

Large inclusions act as massive internal stress concentrators.

The effect is particularly severe in the transverse direction of wrought products compared to the longitudinal direction.

So for really high reliability parts, you need really clean metal.

Exactly.

High purity techniques like vacuum melting are often necessary to reduce these inclusions and meet those demanding fatigue requirements.

Given all these failure mechanisms, engineers have evolved four distinct design philosophies to manage fatigue risk.

And they really reflect a shift from highly conservative pre -crack analysis to managing existing damage.

Right.

The first is infinite life design.

This is the most conservative approach, and it aims for no crack initiation at all.

So all stresses have to be below the fatigue limit.

All stresses, after applying all your correction factors, must be kept below the corrected fatigue limit, sigma e.

This is suitable for components with predictable, low -stress long -term cycling like large stationary machines.

Okay, philosophy number two.

Safe life design.

This assumes the part is initially flawless,

but acknowledges that it has a finite life dictated by initiation and propagation.

And you predict that life?

You predict it based on statistical failure rates, say 99 .9 % survival probability.

And then you retire the part after a set period of service, whether it looks fine or not.

This was historically common in early aircraft design.

The third approach is a step toward redundancy.

Fail safe design.

This ensures the structure has multiple load paths like a bolted assembly.

The idea is that failure of one component is designed to be preceded by large detectable cracks,

allowing the remaining components to carry the load long enough for inspection and repair before catastrophic failure.

And finally, the state -of -the -art approach.

Damage tolerant design.

This is widely used in modern aerospace.

It assumes that flaws or cracks exist from the very beginning.

A pessimistic but realistic view.

A realistic view.

It relies on fracture mechanics, so Paris law, to predict how those cracks will grow under service loading.

This requires periodic non -destructive evaluation, or NDE, to ensure that any detected cracks remain below the critical size AF at all times.

For the conservative everyday approach, the infinite life machine design approach, we mentioned correcting the ideal lab data.

Let's just detail that process one more time.

Okay.

To take the clean laboratory fatigue limit, sigma E prime, and turn it into a realistic working fatigue limit, sigma E, for a real component, we apply a series of empirical correction factors.

So the formula is?

The actual working fatigue limit, sigma E, is calculated as sigma E prime times CS times CF times CZ.

And those C factors are?

CS is the size factor for the reduced strength of large components.

CF is the surface finish factor for surface roughness.

And CZ is the statistical scatter factor, which the engineer selects based on the required reliability, like 99 .9%, to ensure high confidence in the design.

Now let's return to the local strain approach for analysis.

We established that LCF often involves plastic strain at notches.

Even if the nominal stress in the bulk material is elastic,

we need a way to accurately analyze the severe stress and strain right at the notch tip.

Right.

When plastic strain occurs locally, we combine our strain life methodology with a fundamental elasticity concept called Neuber's rule.

What did Neuber propose?

Neuber proposed that the theoretical stress concentration factor, K sigma, is equal to the geometric mean of the local stress concentration factor, K sigma, and the local strain concentration factor, K epsilon.

So KT equals the square root of K sigma times K epsilon.

And this relationship can be rearranged into what's known as the Neuber hyperbola.

Yes.

The relationship is that the product of the local stress range and the local strain range is a constant.

Specifically, the hyperbola is defined by delta sigma times delta epsilon equals the quantity KT times delta S, all squared, divided by E.

Delta S there is the nominal stress range.

And the key to the local strain approach is finding the intersection of this hyperbola with the material's property curve.

Exactly.

You plot the Neuber hyperbola on the stress strain plane.

You also plot the material's stable cyclic stress strain curve, the one we derived from the stable hysteresis loop tips.

And where they cross is your answer.

The coordinates of that single intersection point give you the true local stress range, delta sigma,

and the true local strain range, delta epsilon, occurring right at the critical notch root.

And once you have that accurate local strain, you bypass the nominal stress entirely.

You plug that local strain into the full total strain life equation, and you calculate the precise cycles to failure.

Let's walk through the reasoning for worked example five.

This one analyzes a spline shaft subjected to both an initial high overload and then a subsequent long -term low amplitude operating condition.

This is a complex example because we're analyzing two different loading events, and then we have to sum the damage.

So how do you handle the initial high overload peak?

First, you calculate the nominal stress range for that overload.

Then you apply Neuber's rule, using the KT for this spline groove, to define the Neuber hyperbola equation.

Then find the intersection.

You solve that simultaneous system, the hyperbola and the cyclic stress strain curve, to find the local highly plastic delta epsilon.

You then plug this large delta epsilon into the total strain life equation to calculate the initial short life failure.

Basically, the cycles consumed by that single overload.

Okay, then you tackle the long -term high cycle operating load.

For the long -term lower alternating stress sigma, which is combined with a high mean stress sigma, we find the elastic strain amplitude.

And plug that into the total strain life equation.

Yes, but since there's a high mean stress present, we have to apply a mean stress correction factor, usually a modified Goodman or Soderbergh relationship, to the elastic term.

Why?

This effectively reduces the fatigue strength coefficient sigma's prime.

It accounts for the reduced strength under that steady tensile load.

We then calculate the cycles consumed in this second long -term phase, and the final total life of the shaft is estimated by summing the damage from both phases using Meiner's rule.

So it's a realistic representation of combining different analytical tools to solve a complex history.

It's a great example.

Finally, we need to address how the environment and temperature dramatically alter fatigue properties.

These factors can often transform a component that should have infinite life into one with a very short, finite life.

The most critical environmental factor is corrosion fatigue.

This is the simultaneous effect of cyclic stress and chemical attack.

And even mildly corrosive environments like plain water can have a huge effect.

A dramatic effect.

Corrosion pits act as aggressive stress concentrators, initiating cracks early.

More importantly, the chemical attack at the crack tip prevents the reformation of the protective oxide layer, which aids in crack propagation.

And the most worrying outcome of corrosion fatigue.

What's the bottom line?

It eliminates the conventional fatigue limit.

It's gone.

Completely gone.

Because the corrosive environment keeps driving the crack forward, the SN curve never flattens out, even at extremely low stress levels.

Designing for infinite life in a corrosive environment is impossible.

You have to use a finite life and a damage -tolerant approach.

Another critical surface issue is fretting.

Fletting is surface damage caused by slight, repeated relative motion.

We're talking microns between contacting surfaces.

Like in a bolted join or an interference fit?

Exactly.

This motion causes localized wear, pitting, and the generation of oxide debris, like iron oxide dust.

The pitting and the debris act as strong fatigue crack initiation sites, significantly reducing life, often far below what you'd predict.

Lastly, we have to address temperature effects.

At low temperatures, metals can undergo a ductile -to -brittle transition, which increases their susceptibility to brittle fracture.

However, generally speaking, reducing the temperature often slightly increases fatigue life as long as you avoid that transition to extreme brittleness.

And the opposite is true at high temperatures, where we introduce the complication of creep.

At hot temperatures, fatigue resistance decreases significantly, and the loading frequency becomes critical.

Why frequency?

At high frequencies, fatigue dominates, and you typically get transgranular failure cracks going through the grains.

But as the frequency drops, with slow recycling, the material spends more time under stress at high temperature.

This allows time -dependent creep mechanisms to dominate.

And that changes the failure mechanism.

It shifts the failure mechanism to intergranular cracking, which is cracking along the grain boundaries.

And life is significantly reduced due to this combined creep fatigue interaction.

And finally, the phenomenon of thermal fatigue itself.

This is a classic LCF problem.

Thermal fatigue occurs when fluctuating temperatures generate mechanical stresses, purely because the material is constrained from expanding or contracting.

And since the stress is directly proportional to the temperature change.

The equation is sigma equals alpha e delta t.

This form of loading is entirely strain controlled, which makes the strain life approach the only valid methodology.

This has been an intensive exploration of how materials fail under the relentless march of cycles.

To quickly summarize, we've really dissected three main pillars used in modern fatigue analysis, depending on the failure regime.

Right.

First, you have stress life, the SN curve.

That's the foundation for high cycle fatigue, HCF, and infinite life design, all defined by that endurance limit, SE.

Pillar number two.

Strain life, using the Coffin -Manson and Baskin equations.

This is essential for low cycle fatigue LCF applications where you have plastic strain, often driven by thermal or constrained loading.

And the third.

Fracture mechanics, which is the Paris law.

This is the tool for damage tolerant design, predicting crack propagation rate and remaining life once a macroscopic flaw already exists.

So if you're going to consolidate your understanding down to five core concepts and their mathematical definitions, what are the most critical takeaways?

Okay, first,

the mean stress, sigma m, and the stress ratio.

They are non -negotiable for defining the stress state and moving you beyond idealized lab data.

The SN curve establishes the statistical fatigue limit, SE, which, remember, vanishes in non -ferrous and corrosive environments.

Third.

The total strain life equation.

It's the universal model that ties the shallow elastic part from Baskin and the steep plastic part from Coffin -Manson together.

Fourth.

The Paris law, which governs macroscopic crack growth rate based on the stress intensity factor range, delta K, in that linear region too.

And finally, number five.

The fatigue notch factor, KF.

It's essential.

It links the geometric discontinuities, KT, to the actual observed reduction in fatigue strength, acknowledging the material's ability to resist local high stress through plasticity.

So we've learned how to predict failure under constant amplitude and even simple stepwise variable loading using Miner's rule.

But the greatest complexity in design remains accurately predicting cumulative damage under truly random real -world loading histories, like the turbulence hitting an airplane wing or the random strain on an oil rig.

Right.

Miner's rule offers this simple linear estimate, but it just can't capture the non -linear complex sequence effects where one high load event changes the residual stress field and affects all the subsequent low -load events.

So we absolutely need those more specialized statistical tools.

Particularly the rainflow counting method.

We need that to translate that random irregular data stream into equivalent constant amplitude cycles that are suitable for our life prediction models.

That bridge between the laboratory test and the full -scale component life prediction is still being perfected.

A huge thank you to the Last Minute Lecture team for providing these detailed sources and helping us structure this extensive deep dive.

We hope this exploration provided the clarity and context you needed to master the mechanics of metal fatigue.

Understand those curves,

master those equations, and remember those critical correction factors.

Until next time, stay curious.