What is Vibration Fatigue Analysis?
🙋
What exactly is "vibration fatigue," and why do we use a PSD instead of just looking at a regular vibration signal?
🎓
Great question! Basically, vibration fatigue is the failure of a component from repeated, random shaking—like an engine mount or a circuit board in a rocket. A regular signal shows vibration at each instant, but it's chaotic. A Power Spectral Density (PSD) plot, which you can select in the simulator, is a statistical "recipe" that tells us how the vibration energy is distributed across different frequencies. It's the standard way to define random vibration in engineering tests.
🙋
Wait, really? So the "Modal Stress Coefficient" (cσ) in the tool is like a magic number that turns shaking into stress? How do I get that number?
🎓
Exactly! It's the key link. In practice, you get cσ from a Finite Element Analysis (FEA) modal simulation. You shake your 3D model at its natural frequency with 1 g of acceleration and read the resulting stress. That stress value is your cσ. Try typing a value, say 50 MPa/g, into the simulator. Now, if the PSD says there's vibration at that frequency, the tool uses cσ to calculate the stress it causes.
🙋
Okay, I see the "Steinberg 3σ Method" gives a peak stress. Why multiply by 3? And what does that tell me about when the part will break?
🎓
In practice, for random vibration, the stress levels are statistically distributed. The "3σ" (three-sigma) rule assumes that 99.73% of all stress peaks will fall within ±3 times the RMS (average) stress. So, $\sigma_{peak}= 3 \times \sigma_{rms}$ is a conservative design peak. The tool then uses this peak stress with your material's S-N curve data (the exponent *b* and reference life) to predict fatigue life. Slide the damping ratio (ζ) up and down—you'll see how more damping lowers the stress and extends the life!
Physical Model & Key Equations
The core of the calculation is finding the root-mean-square (RMS) stress response of a structure subjected to random vibration defined by a PSD. The response is dominated by the resonance at the natural frequency.
$$
\sigma_{rms}= c_\sigma \cdot \sqrt{\int_{f_1}^{f_2}W_{resp}(f) \, df}$$
Here, $c_\sigma$ is the modal stress coefficient (MPa/g). $W_{resp}(f)$ is the response PSD, which is the input PSD $W_0$ amplified by the system's transmissibility around the natural frequency $f_n$. The integral calculates the total mean-square acceleration, which $c_\sigma$ converts to stress.
The Steinberg method uses the 3σ peak stress to perform a simplified fatigue damage calculation, assuming the stress follows a Gaussian distribution and breaks the life prediction into three stress levels.
$$
D = f_n \cdot T \sum_{i=1}^{3}\frac{p_i}{N(\sigma_i)}\quad \text{where}\quad \sigma_1 = 1\sigma_{rms},\; \sigma_2 = 2\sigma_{rms},\; \sigma_3 = 3\sigma_{rms}$$
$D$ is the accumulated damage (failure when $D \geq 1$). $f_n$ is the natural frequency (Hz), $T$ is the time duration, $p_i$ is the probability fraction of time spent at each stress level (68.3%, 27.1%, 4.33% respectively), and $N(\sigma_i)$ is the allowable cycles from the S-N curve: $N_i = N_{ref}(\sigma_{ref} / \sigma_i)^b$.
Real-World Applications
Automotive Electronics: Engine control units (ECUs) and sensors are mounted on the chassis or engine and experience road-induced vibration. Engineers use this exact method to specify the required fastener strength and PCB solder joint quality to survive the vehicle's warranty period, often running analyses for millions of simulated road miles.
Aerospace Satellite Components: During rocket launch, payloads endure extreme random vibration. Every bracket, antenna, and circuit board inside a satellite is analyzed for vibration fatigue. A common case is using the PSD levels from the launch vehicle's specification to ensure solar panel deployment mechanisms won't fail before reaching orbit.
Industrial Machinery Piping: Pumps, compressors, and turbines cause high-frequency vibration in connected pipes and supports. Fatigue analysis prevents catastrophic leaks or cracks in chemical plants or power generation facilities. The modal stress coefficient (cσ) is often found by simulating a vibration mode where the pipe bends.
Consumer Electronics Durability Testing: When a company like a smartphone manufacturer tests for "ruggedness," they subject devices to standardized random vibration profiles. This simulation helps predict if internal welds, screen adhesives, or chip connections will fail from being carried in a moving vehicle or during shipping.
Common Misconceptions and Points to Note
When you start using this method, there are a few common pitfalls. First, the point: "The PSD input is acceleration, but is the conversion factor to stress, cσ, really constant?" The modal stress coefficient, cσ, is constant only within the range where a linear, proportional relationship holds. For instance, if large deformations introduce geometric nonlinearities or the material enters the plastic region, cσ will change. In practice, it's a golden rule to run FEA at the anticipated vibration levels to verify that the relationship between stress and acceleration is truly linear.
Next, a fundamental question: "Is the single-degree-of-freedom (single mode) approximation really sufficient?" Certainly, if the component's resonant frequencies are widely spaced and there is only one dominant mode, this is fine. However, if two modes are close together (e.g., 100Hz and 110Hz), they can interact, causing "modal coupling" where the RMS stress becomes larger than a simple summation. If the tool calculates an extremely short or long fatigue life, suspect modal overlap.
Finally, note that "Steinberg's three-band method is not a universal solution." Accuracy can degrade if the slope (exponent b) of the material's S-N curve is very large or small, or if the vibration has strong non-Gaussian characteristics (e.g., containing many shocks). Always remember it is a "simplified, conservative estimation method." For critical designs, it's advisable to compare results with more refined probability density function-based methods like the Dirlik method.
Worked Example
Cast aluminum compressor bracket, fn=62 Hz, ζ=0.04. PSD input: 0.006 g²/Hz (50–100 Hz), 0.012 g²/Hz (100–200 Hz), 0.004 g²/Hz (200–500 Hz). Material: E=72 GPa, allowable stress 180 MPa, Kt=1.5. Calculator yields RMS acceleration 3.8 g, RMS stress 94 MPa, peak stress (3σ)=282 MPa, equivalent static load 8240 N. Using S-N curve for aluminum (slope b=8), fatigue damage D=0.34 over 1000 operating hours, predicting safe operation with 2940 hours to failure margin.