Soderberg Criterion: A Conservative Approach to Fatigue Design

Most mechanical failures aren't caused by a single, catastrophic overload, but by fatigue—the gradual damage inflicted by repeated stress cycles. For engineers designing everything from engine components to aircraft structures, preventing fatigue failure is a paramount concern. The challenge lies in creating parts that can safely withstand not just a steady load, but also the constant vibrations of real-world operation. This requires a robust framework for predicting failure under complex loading conditions, where both static and dynamic forces are at play.

This article addresses this fundamental design problem by exploring fatigue failure criteria. We will dissect cyclic loading into its two key ingredients: mean stress (the steady load) and alternating stress (the cyclic component). You will learn how these stresses are used to map out a component's safety on a Haigh Diagram and how different engineering philosophies lead to different safety boundaries. The core issue is deciding where to draw the line between a lifetime of safe operation and eventual failure.

We will focus on the Soderberg criterion, one of the most important and characteristically conservative models in fatigue design. The following chapters will first explain the core "Principles and Mechanisms," detailing how the Soderberg line is derived from yield strength and endurance limit, and contrasting it with other models like Goodman and Gerber. Subsequently, the "Applications and Interdisciplinary Connections" chapter will explore how this theoretical model is applied in real-world engineering, considering factors like stress concentrations and its relationship with materials science and other disciplines.

Imagine you have a metal paperclip. If you pull on it with all your might, you might bend it, or if you’re incredibly strong, you might even break it. This is a story of static failure—a single, overwhelming event. But we all know there’s another, more insidious way to break that paperclip: bend it back and forth, again and again. It doesn’t take nearly as much force on any single bend, but the repetition, the cycling, eventually causes the metal to snap. This is the world of ​​fatigue​​, and it’s how the vast majority of machine parts, from airplane wings to engine crankshafts, ultimately meet their end.

Our journey is to understand how to design things that don’t fail this way. To do that, we need a map, a guide that tells us what combinations of loading are safe and what combinations lead to disaster.

Let’s look more closely at that wiggling motion. Any cyclic stress, no matter how complicated, can be broken down into two simple components. Think of a wave in the ocean. It has a certain height from trough to peak, but it also sits on an average sea level.

In our world of materials, the "average sea level" is the ​​mean stress​​, denoted by $\sigma_m$. It’s the steady, constant load that the part is under. The "height of the wave" is the ​​alternating stress​​ (or stress amplitude), $\sigma_a$. It’s the part of the load that varies cyclically, the wiggle itself. Mathematically, if a part cycles between a maximum stress $\sigma_{\max}$ and a minimum stress $\sigma_{\min}$, we define them as:

As it turns out, fatigue life depends critically on both of these numbers. It’s not just how big the wiggle is, but also whether the wiggle is happening on top of a big steady pull (a high tensile mean stress) or a steady squeeze (a compressive mean stress). A tensile mean stress is like pulling a guitar string taut before you pluck it—it’s "pre-tensioned" and generally more vulnerable to fatigue damage.

To make sense of this, engineers have created a wonderful tool: the ​​Haigh Diagram​​. It’s a simple chart where we plot the mean stress $\sigma_m$ on the horizontal axis and the alternating stress $\sigma_a$ on the vertical axis. Every possible cyclic loading condition for a component is a single point on this map.

Our mission is to draw a boundary on this map. Inside the boundary is the "safe zone," where a component can withstand its cyclic load indefinitely (or for a very, very long time). Outside the boundary is the "danger zone," where fatigue failure is expected. But how do we draw this line? We start with two known landmarks.

​​Landmark 1: The Pure Wiggle Limit.​​ What happens if there is no mean stress ($\sigma_m = 0$)? We have a purely back-and-forth, symmetric loading. Through experiments, we find that for many materials (like steel), there is a magical threshold for the alternating stress. Below this threshold, called the ​​endurance limit​​ ($S_e$ or $\sigma_e$), the material can be cycled forever without failing. This gives us our first point on the map: on the vertical axis, at $(0, S_e)$. Any load below this point is safe, so long as there's no mean stress.

​​Landmark 2: The No Wiggle Limit.​​ Now, what happens at the other extreme, with no alternating stress ($\sigma_a = 0$)? This is just a static pull. Here, we face a philosophical choice. When do we say a part has "failed"? Does it fail when it starts to permanently bend, an event called ​​yielding​​ which happens at the ​​yield strength​​ ($S_y$ or $\sigma_y$)? Or does it fail only when it snaps in two, which happens at the ​​ultimate tensile strength​​ ($S_u$ or $\sigma_u$)? For any ductile metal, yielding happens long before it snaps, so $\sigma_y \sigma_u$.

This choice gives us two possible landmarks on the horizontal axis: the cautious point $(\sigma_y, 0)$ or the more daring point $(\sigma_u, 0)$.

This brings us to the heart of the matter. How do we connect our landmark on the vertical axis, $(0, S_e)$, to our chosen landmark on the horizontal axis? The simplest thing to do is draw a straight line. This simple act of geometry represents a profound statement about our design philosophy.

Let’s meet the three most famous philosophers of fatigue design: Gerber, Goodman, and our protagonist, Soderberg.

• The ​​Gerber criterion​​ uses a more optimistic parabola connecting $(0, S_e)$ and $(\sigma_u, 0)$. It often fits experimental data for ductile steels well.
• The ​​Goodman criterion​​ takes a more pragmatic approach, drawing a straight line between $(0, S_e)$ and $(\sigma_u, 0)$.
• And then there is the ​​Soderberg criterion​​. The Soderberg philosophy is one of ultimate caution. It declares that any permanent deformation is a failure. A part that has yielded may no longer fit correctly, maintain its alignment, or hold the protective residual stresses it was designed with. For the Soderberg engineer, this is unacceptable. Therefore, the Soderberg criterion draws its straight line from the endurance limit $(0, S_e)$ to the yield strength $(\sigma_y, 0)$.

The equation for this straight line, the ​​Soderberg criterion​​, is beautifully simple:

Any combination of $(\sigma_m, \sigma_a)$ that satisfies this equation lies on the failure line. To be safe, our design point must lie underneath it.

Because $\sigma_y$ is always less than $\sigma_u$ for a ductile material, the Soderberg line is always more restrictive than the Goodman line. It carves out the smallest safe operating region on our Haigh diagram. It is, by design, the most conservative of the common criteria.

Is the Soderberg criterion just a case of engineering paranoia? Absolutely not. It is a deliberate and wise choice for situations where the consequences of yielding are severe.

Consider a high-precision machine tool. If a component yields, even slightly, the machine could lose its accuracy, ruining every part it makes thereafter. Or think of an aircraft landing gear. A design based on the Goodman criterion might be perfectly safe from snapping in two during normal operation. But what if a hard landing—a rare overload event—causes a stress spike that exceeds the yield strength? The landing gear might permanently bend. It hasn't "broken," but it's no longer functional or safe. The Soderberg criterion provides a built-in defense against this kind of failure by ensuring the peak stress in any cycle, $\sigma_{\max} = \sigma_m + \sigma_a$, never exceeds the yield strength.

The gap between the Goodman and Soderberg predictions is itself a fascinating story. It depends on the material's ability to ​​strain harden​​—that is, to get stronger after it starts to yield. For a material with a large gap between its yield and ultimate strengths (a low $\sigma_y / \sigma_u$ ratio), the difference between the Soderberg and Goodman lines is enormous. The choice of philosophy has huge consequences. For a material that is more brittle and breaks soon after it yields ($\sigma_y / \sigma_u \approx 1$), the two lines are nearly identical. The choice matters less. This reveals a beautiful unity: our high-level engineering design philosophy is intimately tied to the fundamental, microscopic behavior of the material itself.

What if we have a compressive mean stress ($\sigma_m 0$)? A steady squeeze should help, right? It should tend to hold any microscopic cracks closed, making it harder for them to grow. Physically, this is correct.

If we naively extend the Soderberg or Goodman lines into the compressive region, they predict that we can handle a larger and larger alternating stress. But here, our cautious engineering sense must kick in again. Blindly trusting the formula could lead us to predict an infinite strength that doesn't exist. More importantly, we risk forgetting about another failure mode entirely. A huge alternating stress, even on top of a compressive mean, could lead to the part being crushed on its minimum-stress swing (compressive yielding) or buckling if it's a slender part.

So, in practice, engineers treat the compressive region with care. Often, they conservatively assume no benefit from compressive mean stress, capping the allowable alternating stress at the endurance limit $S_e$. Or, if they do account for a benefit, they add a separate, explicit check to protect against compressive yielding. This is a profound lesson: engineering is not just about applying equations; it's about understanding the physics behind them and, crucially, the limits of their applicability.

We have spent some time getting to know the Soderberg criterion—a straight line on a map, a simple equation. But a map is only useful if it guides you through a real landscape. Now, we leave the clean, abstract world of theory and venture into the messy, vibrant, and fascinating world of engineering practice. What is this tool really for? How does it help us build things that work, and more importantly, things that don't fail? You will see that this simple line is not just a formula; it is a philosophy of design, a statement of caution, and a bridge connecting the mechanics of materials to a host of other scientific disciplines.

Imagine you are tasked with designing a steel tie-rod for a machine that will hum along for millions of cycles. You have your fatigue data, and you're ready to design for "infinite" life. The Goodman and Gerber criteria, as we've seen, offer generous operating envelopes by anchoring their limits to the material's ultimate tensile strength, $S_u$. They whisper, "As long as you stay within this boundary, fatigue won't get you." But there is a catch, and it's a big one. What good is a component designed to last forever against the gentle whisper of fatigue if it plastically deforms like a piece of taffy under the first significant load?

This is the fundamental question that the Soderberg criterion forces us to confront. It is the most conservative of the common fatigue models precisely because it chooses a different anchor point for static failure: the yield strength, $S_y$. Its guiding principle is not merely to prevent a fatigue crack from forming over a million cycles, but to ensure the component doesn’t surrender by yielding on the very first cycle. By drawing its boundary line from the endurance limit, $S_e$, to the yield strength, $S_y$, the Soderberg criterion creates a safe harbor where neither fatigue nor first-cycle yielding can occur.

While the Goodman or Gerber criteria might permit a design where the maximum stress, $\sigma_{\max} = \sigma_m + \sigma_a$, exceeds the yield strength, the Soderberg criterion strictly forbids this. For ductile materials, where the gap between yield and ultimate strength can be large, this difference is profound. The Soderberg approach embodies a philosophy of absolute elastic design; it ensures that under the most extreme point in any load cycle, the component remains perfectly elastic. This conservatism makes it a favorite in applications where any amount of plastic deformation is unacceptable, such as in precision machinery or structures where dimensional stability is paramount. Incorporating a safety factor, $n$, further shrinks this safe harbor, creating an even larger buffer between the operating stresses and the material's limits, ensuring a robust and reliable design.

Our neat diagrams often assume a polished, uniform bar of metal. But real-world parts are not so simple. They have holes for bolts, grooves for snap rings, and fillets where shafts change diameter. Each of these geometric features is a-crack-in-waiting. Anyone who has ever torn a piece of paper knows that the tear starts much more easily if you first make a tiny nick. The same principle applies to metals.

These geometric features act as "stress raisers," causing the local stress in their immediate vicinity to be far higher than the nominal stress you would calculate by simply dividing the load by the cross-sectional area. This is the tyranny of geometry. We quantify this effect with a fatigue stress concentration factor, $K_f$, which accounts for both the theoretical sharpness of the notch ($K_t$) and the material's sensitivity to that notch ($q$). When we apply the Soderberg criterion to a real part, we must use the local stresses in our calculation: $\sigma_{a, \text{local}} = K_f \sigma_a$ and $\sigma_{m, \text{local}} = K_f \sigma_m$. Suddenly, our safe operating stress is much lower than we thought! This crucial step connects the abstract failure line to the tangible, physical shape of the component.

But if geometry can be an enemy, we can also use material science to make it an ally. Many high-performance components, like engine crankshafts or aircraft landing gear, undergo surface treatments such as shot peening or case hardening. These processes create a "pre-loaded" surface layer with a high compressive residual stress. Imagine trying to pull apart a book while your friend is squeezing it shut; the squeeze provides a built-in resistance. This compressive stress, $\sigma_r$, adds algebraically to the applied mean stress, $\sigma_m$. The total effective mean stress the material actually experiences is $\sigma_{m, \text{eff}} = \sigma_m + \sigma_r$. Since the residual stress is compressive ($\sigma_r 0$), it can turn a damaging tensile mean stress into a harmless, or even beneficial, compressive one. This is a beautiful example of engineering at its finest—fighting fatigue by building a defense directly into the material's surface.

The principles of fatigue are not confined to the world of mechanical engineering. They are deeply interwoven with materials science, chemistry, and thermodynamics.

What happens when we design with a material, like an aluminum alloy, that doesn't have a true endurance limit? Its strength continues to decrease with more cycles, no matter how small the stress. Here, the idea of "infinite life" becomes meaningless. Must we abandon our models? Not at all. We simply adapt. Instead of an endurance limit, $S_e$, we use a finite-life fatigue strength, $S_f$, which is the stress amplitude the material can survive for a specific target number of cycles (say, $10^7$). The Soderberg, Goodman, and Gerber diagrams are then redrawn as constant-life diagrams for that target life, allowing us to design for a finite, but known, service period. This connects our mechanical model to the fundamental stress-life behavior dictated by the material's microstructure.

Now, let's turn up the heat. When a steel component operates at elevated temperatures, its atoms vibrate more vigorously. This thermal energy helps dislocations—the tiny defects that govern plastic flow—to move and overcome barriers more easily. As a result, the material softens: its yield strength, ultimate strength, and endurance limit all decrease. On our fatigue diagram, this is catastrophic. The entire Soderberg failure line shrinks and moves inward, toward the origin. The safe operating envelope becomes drastically smaller. An alternating stress that was perfectly safe at room temperature might now lead to rapid failure. This is a direct link between mechanical strength and thermodynamics.

The environment can be an even more insidious enemy. Operating a steel shaft in a saltwater environment is a recipe for corrosion fatigue. The chemical attack from the corrosive medium does two terrible things. First, it makes it easier for fatigue cracks to start, drastically reducing the material's endurance strength. But second, and more subtly, it can make the material more sensitive to mean stress. A tensile mean stress, which pulls the crack faces apart, allows the corrosive fluid to penetrate deeper and attack the highly stressed material at the crack tip, accelerating its growth. A less-conservative model like the Gerber parabola, which often works well for ductile metals in air, can become dangerously non-conservative in a corrosive environment. The increased sensitivity to mean stress pushes us back toward a more cautious linear model, like Goodman or, for maximum safety, Soderberg. Here, our choice of mechanical model is dictated by chemistry.

We have seen the power and versatility of the Soderberg criterion. It gives us a robust framework for designing components that must withstand millions of stress cycles. But as with any model, its greatest strength comes from knowing its limitations. The entire stress-life philosophy, which includes Soderberg, Goodman, and Gerber, is fundamentally about predicting crack initiation. It assumes we are starting with a clean, smooth, nearly flawless material.

What happens if a flaw is already present? All manufactured components contain tiny defects—microscopic voids, inclusions from the manufacturing process, or tiny scratches from handling. If one of these defects is large enough, the game changes completely. The question is no longer "How long until a crack starts?" but "How long until this crack grows to a critical size and the part breaks?"

This is the domain of ​​Linear Elastic Fracture Mechanics (LEFM)​​. Consider a high-strength steel plate where our Soderberg analysis predicts infinite life for the given service stress. Now, suppose an inspection reveals a tiny, pre-existing surface crack. LEFM allows us to calculate the stress intensity at the tip of that crack. If this intensity exceeds a certain material threshold, the crack will grow with every load cycle, regardless of what our S-N diagram says. A component predicted to last forever might, in fact, have a finite life of a few hundred thousand cycles.

This realization is the boundary of our map. The Soderberg criterion is the perfect tool for an "initiation-controlled" design philosophy. But for safety-critical components made of high-strength, low-toughness materials, or where we know initial flaws are unavoidable, we must switch to a "damage-tolerant" philosophy and use the more sophisticated tools of fracture mechanics. Knowing when to make this switch—when to trade our trusted map for a more powerful satellite navigation system—is the hallmark of a wise engineer. The Soderberg criterion is not the final word on fatigue, but it is an essential, eloquent, and profoundly useful first chapter.