Gerber Criterion

The silent, progressive failure of materials under fluctuating loads, known as fatigue, is a primary concern in virtually every field of engineering. While it's simple to test a material's strength under a steady pull, real-world components from bridges to engine shafts are subjected to a complex mix of steady (mean) stress and vibration (alternating stress). This raises a critical question: how do these stresses combine to cause failure? Predicting this interaction accurately is essential for designing safe, reliable, and efficient structures. This article tackles this knowledge gap by providing a deep dive into one of the most effective tools for this task: the Gerber criterion.

Over the next sections, you will embark on a journey from foundational theory to practical application. The first chapter, ​​"Principles and Mechanisms,"​​ will introduce the core concepts of fatigue, contrast the simple linear Goodman model with the more physically-grounded Gerber parabola, and use experimental data to demonstrate the Gerber criterion's superior accuracy for ductile materials. The second chapter, ​​"Applications and Interdisciplinary Connections,"​​ will move from theory to practice, exploring how engineers use the Gerber criterion in design, the crucial role of stress concentrations, the powerful technique of pre-stressing, and most importantly, the boundaries of the model and its relationship to the separate field of fracture mechanics.

Imagine you have a paperclip. If you pull on it with a steady force, it takes quite a bit of effort to break it. But if you bend it back and forth a few times—applying a fluctuating, or ​​alternating stress​​—it snaps easily. This phenomenon, known as ​​fatigue​​, is the silent enemy of almost every machine and structure we build, from airplane wings to bridges to the engine parts in your car. The core question for any engineer is: how much of this "wiggling" can a part withstand before it fails?

Let’s try to map out this problem. We can define our "wiggling" by its amplitude, which we'll call the ​​alternating stress​​, $\sigma_a$. We can also have a steady, average pull on the part, which we call the ​​mean stress​​, $\sigma_m$.

We know two facts for sure. First, if there is no steady pull at all ($\sigma_m = 0$), a material can withstand a certain maximum wiggle amplitude indefinitely without failing. This magical threshold is called the ​​endurance limit​​, $S_e$. It is our first landmark on the map of safety. Second, if there is no wiggling at all ($\sigma_a = 0$), the material will fail like a simple rope if we pull on it hard enough, at a stress called the ​​ultimate tensile strength​​, $S_u$. This is our second landmark.

Our map, often called a ​​Haigh diagram​​, plots mean stress $\sigma_m$ on the horizontal axis and alternating stress $\sigma_a$ on the vertical axis. We have two points on the boundary between "safe" and "failure": $(0, S_e)$ and $(S_u, 0)$. The million-dollar question is: what does the boundary line connecting these two points look like? This line will tell us, for any given mean stress $\sigma_m$, what is the maximum allowable alternating stress $\sigma_a$.

What's the simplest way to connect two points? A straight line, of course! This was the sensible suggestion made by Goodman. The ​​Goodman criterion​​ simply draws a straight line between our two landmarks. Mathematically, this line is described by a beautifully simple equation:

This equation says that the fraction of the "wiggle-life" we use up ($\sigma_a/S_e$) plus the fraction of the "static-pull-life" we use up ($\sigma_m/S_u$) must not exceed one. It implies that for every bit of steady pull we add, we must give up a proportional amount of wiggle room. It’s a very reasonable, tidy, and often used rule. As we’ll see, it's also quite cautious, providing a good margin of safety.

But is nature always so simple? Does a tiny bit of mean stress immediately start "eating away" at our wiggle room at a constant rate? Let's think about the physics. For a tough, ​​ductile​​ material like steel, you might imagine that it can shrug off a small amount of mean stress. Its effect might be negligible at first, only becoming significant as the mean stress gets larger.

If the effect is negligible at the start (at $\sigma_m = 0$), it means the slope of our failure boundary on the Haigh diagram should be horizontal, or zero, at that point. A straight line can't do that; its slope is constant. So, what's the next simplest curve that can start flat and then curve downwards to meet our other landmark at $(S_u, 0)$? The answer is a parabola.

This elegant bit of reasoning leads us to the ​​Gerber criterion​​. It keeps the same two landmarks but connects them with a parabola. The equation is just as lovely, but with one crucial change: the mean stress term is squared.

Notice the geometry. Because $(\sigma_m/S_u)^2$ is smaller than $(\sigma_m/S_u)$ for any pull less than the ultimate strength, the Gerber parabola bows upwards, arching above the straight Goodman line. This means that for a given mean stress $\sigma_m$, the Gerber criterion predicts that the material can handle a higher alternating stress $\sigma_a$. It is less conservative than the Goodman line, suggesting that ductile materials are tougher against combinations of mean and alternating stress than a simple linear guess would predict.

This is all very nice armchair philosophy, but in science, the ultimate judge is experiment. Does the Gerber parabola actually describe how real materials behave? Let’s run a hypothetical test in our lab.

Imagine we are testing a high-strength nickel alloy with an endurance limit $S_e = 450 \text{ MPa}$ and an ultimate strength $S_u = 900 \text{ MPa}$. We apply a steady mean stress of $\sigma_m = 540 \text{ MPa}$ and start wiggling it, increasing the amplitude $\sigma_a$ until it fails after a million cycles. We find that failure occurs at an amplitude of about $\sigma_a \approx 285 \text{ MPa}$.

Now, what did our models predict?

• ​​Goodman (linear):​​ $\sigma_a = 450 \left(1 - \frac{540}{900}\right) = 450(1 - 0.6) = 180 \text{ MPa}$
• ​​Gerber (parabolic):​​ $\sigma_a = 450 \left(1 - \left(\frac{540}{900}\right)^2\right) = 450(1 - 0.36) = 288 \text{ MPa}$

The result is striking. The Goodman line was far too pessimistic, predicting failure at a much lower amplitude than what we observed. The Gerber parabola, however, predicted the failure amplitude with stunning accuracy. Similar tests on other ductile materials, like certain aluminum alloys, show the same thing: the data points for failure consistently fall much closer to the Gerber parabola than the Goodman line. This is a powerful lesson: a simple, physically motivated refinement to our model—the idea of a gentle start—led to a formula that beautifully captures the true behavior of many materials.

The success of the $(\sigma_m)^2$ term beckons a deeper question: is there a physical reason for the square? Why isn't it $\sigma_m^{1.9}$ or $\sigma_m^{2.1}$?

To get a hint, we must zoom into the microscopic world of the material. Fatigue failure is the story of microscopic cracks growing over millions of cycles. A tensile mean stress helps this process by pulling the crack faces apart. Now, let’s imagine a "damage" mechanism at this tiny scale. If the fundamental process is symmetric—that is, it doesn't initially care if it's being 'opened' by a small positive or negative mean stress—then the function describing it must be an even function. The simplest nontrivial even function is a quadratic, like $x^2$.

So, we can postulate a "damage" variable, $D$, that degrades the material's ability to resist wiggling, and that for small mean stresses, this damage is proportional to $\sigma_m^2$. The allowable alternating stress would then be reduced by this damage: $\sigma_a = S_e(1 - \text{const} \cdot \sigma_m^2)$. This is precisely the form of the Gerber relation! The ultimate strength, $S_u$, then enters the picture as a convenient, experimentally measured parameter to scale the equation so that it lands on the correct second landmark $(S_u, 0)$. In essence, the Gerber parabola isn't just a random curve that fits the data; its quadratic form can be seen as a natural consequence of a symmetric, energy-like damage process at the heart of the material.

A good scientist, and a good engineer, knows not only how a tool works but also when it doesn't work. The Gerber criterion is powerful, but it is not a universal law.

​​The Peril of Brittleness​​: The Gerber parabola's optimism is rooted in the toughness and ductility of the material. But what if our material is more brittle, or behaves in a brittle way because of its heat treatment or the specific loading conditions? For such materials, a tensile mean stress can be far more detrimental. They don't have that initial "shrug it off" quality. In these cases, the failure boundary lies much closer to the conservative Goodman line. An engineer who blindly applies the Gerber formula to a brittle material might design a part that is dangerously unsafe, as it would fail under loads the formula predicted it could handle.

​​The Compressive Friend​​: What happens when the mean stress is compressive ($\sigma_m 0$)? A steady push should tend to close micro-cracks, hindering their growth. So, a compressive mean stress should be beneficial, allowing for an even larger wiggle amplitude $\sigma_a$ than the endurance limit $S_e$. However, if you look at the Gerber formula, with its $\sigma_m^2$ term, it predicts the exact same reduction in $\sigma_a$ for a compressive stress as for a tensile one! This is physically wrong. It's a stark reminder that we must not extrapolate mathematical formulas beyond the physical regime where they were validated. In practice, engineers handle compressive mean stress very carefully. A common conservative approach is to simply ignore any benefit and cap the allowable alternating stress at $S_e$ for any $\sigma_m \le 0$.

​​The Ultimate Safety Net​​: Finally, there are situations where we must be supremely cautious. What if we cannot allow the part to permanently deform (or ​​yield​​) even slightly? In this case, we must ensure that the peak stress in any cycle, $\sigma_{\text{max}} = \sigma_m + \sigma_a$, never exceeds the material's ​​yield strength​​, $S_y$. This leads to an even more conservative linear rule called the ​​Soderberg criterion​​, which connects $(0, S_e)$ to $(S_y, 0)$. Since the yield strength $S_y$ is always less than the ultimate strength $S_u$ for ductile metals, the Soderberg line carves out the smallest safe operating region of all. It's a guarantee against both fatigue and yielding, representing the highest level of design caution.

The journey through these criteria—from a simple line to a subtle curve and its boundaries—reveals the beautiful interplay between physical intuition, mathematical modeling, and experimental validation that lies at the heart of engineering science.

We have spent some time getting to know the Gerber parabola—an elegant, smooth curve that captures how a steady, or mean, stress affects a material's ability to withstand vibrations. We saw that it provides a more optimistic, and often more accurate, prediction of fatigue life for many materials compared to its linear cousins, the Goodman and Soderberg criteria. But a physical law, no matter how elegant, is only as good as what it allows us to do. What is the real-world value of this equation? Where does it leave the pages of a textbook and enter the world of humming engines, soaring aircraft, and spinning turbines?

This is where the real fun begins. We are now moving from the "what" to the "so what." We will see how engineers wield the Gerber criterion not as a mere formula, but as a lens through which to view the complex interplay of stress, geometry, and material history. It becomes a crucial tool in a grand endeavor: designing a world that doesn't fall apart.

Imagine you are an engineer designing a critical tie-rod for a machine. It will be pulled with a steady force, but it will also vibrate back and forth. You know the alternating stress, $\sigma_a$. The question is, what is the maximum steady, mean stress, $\sigma_m$, you can safely allow?

Here, the choice of fatigue criterion becomes a choice of design philosophy. Let's look at three options you might have:

1. ​​The Soderberg Criterion:​​ This is the most conservative approach. It draws a straight line from the material's endurance limit ($S_e$) on the alternating stress axis to its yield strength ($S_y$) on the mean stress axis. It builds in a large safety margin by insisting that the maximum stress in any cycle ($\sigma_m + \sigma_a$) never, ever causes the material to permanently deform or "yield." It is supremely cautious.

2. ​​The Goodman Criterion:​​ This is a more common and realistic approach. It also draws a straight line, but this time to the material's ultimate tensile strength ($S_u$)—the stress at which it breaks. The philosophy here is that fatigue failure is the primary concern, not minor yielding.

3. ​​The Gerber Criterion:​​ This is our star player. Instead of a straight line, it draws a parabola to the same ultimate strength point as Goodman. For the same alternating stress, the Gerber criterion will almost always permit a higher mean stress than Goodman. For a tie-rod designed with an alternating stress of $90 \, \text{MPa}$ and made of a steel with $S_u = 700 \, \text{MPa}$, the Gerber criterion might allow a mean stress of nearly $540 \, \text{MPa}$, whereas the Goodman line would cap you at around $414 \, \text{MPa}$.

At first glance, Gerber looks like the clear winner—it allows for a lighter, more efficient design. But nature teaches us there is no free lunch. The solution from problem contains a profound warning: a design that is "safe" according to the Gerber criterion might still yield on its very first load cycle! In that hypothetical example, the maximum stress reached would be $\sigma_{\max} = \sigma_m + \sigma_a \approx 538 + 90 = 628 \, \text{MPa}$, which is far above the material's yield strength of $520 \, \text{MPa}$. The part wouldn't fail from fatigue, but it would be permanently stretched and ruined before the vibrations even got a chance to do their damage.

This reveals a beautiful point about engineering: it is not about blindly applying one rule. It is about orchestrating a symphony of rules. The engineer must use the Gerber criterion to check for fatigue, but also separately check that the peak stress does not cause yielding. A safe design lives in the space allowed by all relevant failure modes.

One of the most powerful ideas in engineering is that you can dramatically improve a component's performance by introducing "residual" stresses before it ever sees a single service load. Think of it like building an arch with a keystone; the compressive forces holding it together are there by design, making the structure immensely strong. In fatigue, we can do something similar.

Treatments like "shot peening" (blasting a surface with tiny beads) or case hardening create a thin layer of material at the surface that is in a state of compression. This compressive stress, let's call it $\sigma_r$, is a negative value. When we apply an external load, the total stress the material feels is simply the sum of the applied stress and this built-in residual stress.

The magic is in how this affects the mean stress. A cyclic load with a positive (tensile) mean stress $\sigma_m$ is now experienced by the material as an effective mean stress $\sigma_m^{\text{eff}} = \sigma_m + \sigma_r$. Since $\sigma_r$ is negative, the effective mean stress is lower than the applied mean stress!.

Imagine a shaft under a fully reversed load, with an applied mean stress of zero. By shot-peening it, we might introduce a residual stress of, say, $-200 \, \text{MPa}$. Suddenly, the material at the surface is no longer cycling around a mean of zero. It is cycling around a beneficial mean stress of $-200 \, \text{MPa}$. On the Haigh diagram, we have literally shifted our operating point from the vertical axis deep into the "safe" territory to the left.

This is not just a theoretical nicety. The practical benefit can be staggering. Using the Gerber criterion, we can quantify this benefit by calculating an "equivalent fully reversed stress." This is the stress of a simple, zero-mean cycle that would be just as damaging as our complex cycle with residual stress. A cycle with an alternating stress of $300 \, \text{MPa}$ and an effective mean stress of $-200 \, \text{MPa}$ might be equivalent to a simple reversed cycle of only $250 \, \text{MPa}$ (using the Goodman model for illustration). The material behaves as if the load is less severe than it actually is.

And the payoff? A massive increase in life. For a steel component under a certain load, introducing a compressive residual stress of $-200 \, \text{MPa}$ could increase its fatigue life by a factor of 2.35, meaning it lasts more than twice as long!. This isn't just cheating failure; it's a deep and clever manipulation of the internal state of matter to make our creations more robust and reliable.

So far, we have been talking as if our components are perfectly smooth, uniform bars. But look around you: real-world parts have holes, fillets, keyways, and shoulders. They have geometry. And wherever the smooth flow of force is interrupted, stress concentrates. Think of water flowing smoothly in a wide channel, and then being forced around a sharp rock; the water right at the edge of the rock speeds up dramatically. Stress does the same thing.

A simple hole in a plate can triple the stress at its edge. This amplification is captured by a theoretical stress concentration factor, $K_t$. Naively, one might think we should just multiply our applied stresses by $K_t$ and check that against our Gerber criterion. But fatigue is a bit more subtle. Due to microscopic yielding and material grain structure, a material doesn't always "feel" the full, sharp peak of the theoretical stress. This effect is captured by a "notch sensitivity" factor, $q$.

The factor that really matters for fatigue is the fatigue strength reduction factor, $K_f$, which is a blend of the theoretical geometry and the material's sensitivity to it: $K_f = 1 + q(K_t - 1)$. This $K_f$ is what we use to find the local stresses at the root of the notch. Both the mean and alternating components of the stress are amplified: $\sigma_{a, \text{local}} = K_f \sigma_a$ $\sigma_{m, \text{local}} = K_f \sigma_m$

The Gerber criterion is then applied not to the nominal stress in the bulk of the part, but to these much higher local stresses. A component that seems perfectly safe based on its average stress might be dangerously close to failure in one tiny, localized spot. The Gerber criterion, applied locally, becomes the ultimate arbiter, telling us if that tiny spot will be the birthplace of a fatal crack.

The S-N approach to fatigue, including our Gerber criterion, is built on a fundamental philosophy: that the vast majority of a component's life is spent initiating a crack. The models predict how long a seemingly "perfect" surface can endure before a microscopic crack is born and begins its journey. For many applications, this is an excellent and effective way to design.

But what if a crack is already there?

High-strength modern materials, while incredibly strong, can be unforgiving of pre-existing flaws from manufacturing or service. A tiny scratch, a microscopic inclusion, or a welding defect might be present from day one. In this case, the "initiation" phase of life is zero. The entire lifetime of the component is dictated by how fast that initial crack propagates.

This is a complete philosophical shift, moving us from the world of Stress-Life (S-N) to the world of Linear Elastic Fracture Mechanics (LEFM). And this is where we must understand the limits of the Gerber criterion.

Consider a steel plate with a tiny, undetectable surface crack of just $0.25 \, \text{mm}$. Our analysis using the Gerber criterion for the applied loads predicts infinite life—the plate should last forever! We've done our job, and the design is safe. Or is it?

An engineer trained in fracture mechanics would ask a different question: Is the stress fluctuation at the tip of that tiny crack large enough to make it grow? Using LEFM, we can calculate the "stress intensity factor range," $\Delta K$, a measure of the driving force for crack propagation. It turns out that for the given load, the $\Delta K$ for our tiny crack is above the material's threshold for growth, $\Delta K_{th}$.

The conclusion is chilling and profound: the crack will grow a little bit with every single load cycle. The prediction of infinite life from the S-N model is dangerously wrong. By integrating the rate of crack growth (using a model like the Paris law), we can predict the actual finite life—in the scenario of problem, it was on the order of $260,000$ cycles.

This teaches us the most important lesson of all. A model is a map, not the territory itself. The Gerber criterion is a powerful and indispensable map for designing against crack initiation. But for safety-critical components where pre-existing flaws are a major concern (like in aerospace or power generation), we must switch maps. We must adopt a "damage-tolerant" philosophy, assume flaws are present, and use the tools of fracture mechanics to ensure they grow so slowly that the component can be retired or repaired long before they become critical. Knowing when to use the Gerber criterion is just as important as knowing how to use it. It is this wisdom—understanding the domain and limitations of our powerful ideas—that sits at the very heart of great science and engineering.