Reality Check on Random Vibration Simulation
THE ANALYST’S FIELD GUIDE TO
Random Vibration
Understanding the Why Before You Touch the Software
Joseph P. McFadden Sr., Engineering Fellow
Mechanical Engineering Analysis & Services (MEAS)
Zebra Technologies Corporation
1. What Random Vibration Actually Is
1.1 The Physical Meaning of a Random Environment
1.2 Power Spectral Density: The Language of Random Vibration
1.3 How Structures Respond to Random Excitation
2. Test Standards: The Specification Behind the Analysis
2.1 Why the Specification Is Not a Starting Point — It Is a Question
2.2 The Major Standards Landscape
IEC 60068-2-64: Random Vibration
MIL-STD-810, Method 514: Vibration
IEC 60721: Classification of Environmental Conditions
ASTM D4169 and ISTA Protocols: Transportation Vibration
ISO 16750: Road Vehicles — Electrical and Electronic Equipment
2.3 Translating the Specification into Analysis Inputs
3. Understanding the Ask: What the Customer Usually Means
3.1 Compliance Qualification vs. Design Guidance vs. Failure Investigation
3.2 The Questions the Customer Cannot Always Answer — But You Must Ask
3.3 The Ask That Does Not Make Sense Yet
4. Being a Collaborative Partner in the Vibration Space
4.1 Connecting Simulation to the Physical Test Program
4.2 Educating the Customer on What the Analysis Can Say
4.3 The Material Data Gap in Vibration Work
5. Before the First Button Is Pushed: The Path Forward
5.1 The Pre-Analysis Checklist
Problem Definition and Context
Correlation and Validation Plan
5.2 The Modal Analysis Review — Your First Sanity Check
5.3 The Response Analysis and Its Interpretation
5.4 Communicating Results Honestly
6. A Final Word on Physical Intuition
Appendix A: Standards Quick Reference
Appendix B: Pre-Analysis Conversation Template
A Note Before We Begin
Random vibration is one of the most misunderstood analyses in the mechanical engineering toolkit. Not because the physics is obscure — it is well established, grounded in probability theory and structural dynamics that date back nearly a century. The misunderstanding runs deeper than that. It comes from a gap between what the analysis actually describes and what engineers and customers tend to assume it describes.
A drop test simulation produces a result that feels intuitive. You drop the device, something bends or breaks, you see a stress contour and a failure prediction. The cause-and-effect narrative is visible. Random vibration is different. The input is a statistical description of motion spread across a spectrum of frequencies. The output is a statistical description of structural response — probability distributions, RMS stresses, expected fatigue cycles. There is no single, discrete event. There is an environment, and the question is whether the structure can survive sustained exposure to it.
That distinction is important for how you talk to customers, how you set up your models, and how you interpret and communicate your results. An analyst who does not genuinely understand what a power spectral density represents — what it is saying physically, not just mathematically — will produce outputs that may be numerically correct but physically misleading.
This guide is a companion to the drop test guide. It follows the same philosophy: the software is a tool in service of engineering judgment, the most important work happens before the first button is pushed, and your highest obligation is to help your customer make a better decision — not to produce a report.
Random vibration analysis is fundamentally a fatigue and reliability question dressed in the language of structural dynamics. If you approach it as anything else, you will answer the wrong question — even if your numbers are right.
Read this document as a conversation with a colleague who has spent a career at the intersection of vibration analysis, product qualification, and the hard lessons that come from the gap between simulation and physical reality. The goal is to give you a foundation for thinking, not a procedure for clicking.
1. What Random Vibration Actually Is
1.1 The Physical Meaning of a Random Environment
In most engineering loading scenarios we are trained to think about, there is a definable event. A known force, a specified displacement, a particular impact. Random vibration does not work that way. The vibration environment experienced by a product in the field — mounted in a vehicle, carried by a worker, shipped in a package on a truck — is not a single deterministic event. It is a continuous, irregular excitation whose instantaneous value at any moment cannot be predicted, only described statistically.
The key word is stationary. We typically assume that random vibration environments are stationary processes — meaning that their statistical properties do not change over time. The average acceleration is approximately zero. The variance, which characterizes the energy content, remains constant. The frequency content, described by the power spectral density, does not drift. This stationarity assumption is what makes the problem mathematically tractable and is also one of the assumptions you must evaluate for every analysis you perform. Not all real-world vibration environments are truly stationary, and understanding when the assumption breaks down is part of the analyst's job.
A second assumption you will encounter constantly is Gaussianity. Random vibration methods typically assume that the instantaneous acceleration values follow a Gaussian, or normal, probability distribution centered at zero. This assumption allows us to use standard statistical relationships — in particular, the relationship between the RMS value and the probability of exceeding various amplitude thresholds. In practice, many real-world vibration environments are approximately Gaussian over broad frequency ranges, but there are important exceptions: shock-contaminated environments, nonlinear systems, and certain transportation scenarios can all produce non-Gaussian characteristics that require more careful treatment.
Understanding these assumptions is not academic pedantry. It is the difference between knowing what your analysis is actually saying and believing it is saying something it is not.
1.2 Power Spectral Density: The Language of Random Vibration
The power spectral density, or PSD, is the fundamental language of random vibration specification. It describes how the mean-square acceleration of the vibration environment is distributed across the frequency spectrum. The units are typically acceleration squared per unit frequency — g-squared per Hertz in the US, or meters per second squared squared per Hertz in SI.
When you look at a PSD curve, you are looking at a picture of energy distribution. A flat PSD over a frequency range means equal energy per unit bandwidth across that range. A peak in the PSD at a particular frequency means relatively more energy near that frequency. The area under the PSD curve — the integral over the frequency range — equals the mean-square acceleration, whose square root is the overall RMS acceleration level.
This is an important physical checkpoint. When a customer hands you a test specification with a PSD table and a stated overall g RMS level, verify that they are consistent. Integrate the area under the PSD and confirm that the square root matches the stated g RMS. This is a basic sanity check that reveals specification errors more often than you would expect, especially when specifications have been copied and modified from other documents without checking the mathematics.
The PSD also carries information about the nature of the vibration source. Broadband, relatively flat PSDs often represent environments like aircraft turbulence or broad-spectrum road vibration. PSDs with prominent peaks at specific frequencies suggest periodic or tonal content — perhaps from a rotating machine, an engine harmonic, or a resonance in the mounting structure. These spectral shapes matter because your structure will respond differently to different frequency content depending on where its natural frequencies fall. This is the essence of the resonance problem, and it is why modal analysis is always the first step in understanding a structure's vibration behavior.
1.3 How Structures Respond to Random Excitation
When a random vibration environment is applied to a structure, the structure acts as a filter. It amplifies excitation near its natural frequencies and attenuates excitation at other frequencies. The degree of amplification is governed by the quality factor Q, which is the inverse of twice the critical damping ratio. A lightly damped structure — high Q, low damping — will amplify excitation at resonance dramatically. A heavily damped structure will respond more uniformly across the frequency range.
This filtering behavior is captured in the frequency response function, which relates the output response at any point in the structure to the input excitation as a function of frequency. In random vibration, the response PSD at any location is the product of the input PSD and the square of the magnitude of the frequency response function. This is the mathematical foundation of the analysis: find the response PSD, extract RMS stresses, and apply fatigue methods to predict life.
The critical point that many analysts miss — and that almost every customer misses — is that the damage done by random vibration is almost entirely determined by the response at and near the structure's natural frequencies. A product that has a natural frequency inside the input frequency range of the test specification will see vastly more stress than a product whose natural frequencies are outside that range. This is why design decisions that shift natural frequencies, or change damping levels, can have a larger effect on fatigue life than changes to material strength. You cannot evaluate a random vibration result without understanding the modal behavior of the structure.
Random vibration damage is dominated by resonance response. A structure that resonates within the input frequency band will accumulate fatigue damage orders of magnitude faster than one that does not. The modal analysis is not a preliminary step — it is the analysis. Everything else interprets what the modal analysis reveals.
1.4 The Connection to Fatigue
Random vibration analysis is, at its core, a fatigue problem. The structure is subjected to cyclic stress whose amplitude varies randomly. The damage accumulates over time according to Palmgren-Miner's rule — the sum of cycle ratios at each stress amplitude relative to the fatigue life at that amplitude. When the damage sum reaches unity, fatigue failure is predicted.
In random vibration, the stress amplitude distribution follows from the response statistics. For a Gaussian response, the distribution of peaks follows a Rayleigh distribution, and the expected damage can be calculated analytically using the Dirlik method or similar spectral fatigue approaches. These methods transform the frequency-domain response PSD directly into a fatigue damage estimate, without requiring a time-domain stress history.
This is powerful and elegant, but it carries assumptions. The S-N curve you use must be appropriate for the material, the surface condition, the stress concentration at the critical location, and the mean stress state. If you are using published fatigue data without accounting for surface finish, stress concentration, or mean stress effects, your fatigue life prediction may be optimistic by orders of magnitude. This is not a small uncertainty. It is the kind of uncertainty that separates a meaningful analysis from a number that happens to have been calculated.
2. Test Standards: The Specification Behind the Analysis
Every random vibration analysis must be anchored to a physical specification. The specification defines the input environment, the duration, the axes of excitation, the temperature context, and often the acceptance criteria. Without this anchor, you are running a simulation with no connection to the real world, and the results have no engineering meaning regardless of how carefully the model is constructed.
2.1 Why the Specification Is Not a Starting Point — It Is a Question
When a customer hands you a vibration specification, the correct first response is not to start building a model. It is to ask where the specification came from and whether it represents the actual service environment of the product.
Vibration specifications in the industrial electronics world have a long history of being copied from other specifications, rounded up for conservatism, or inherited from predecessor products without verification against current use cases. A handheld device carried by a warehouse worker in a distribution center experiences a very different vibration environment than one mounted in an agricultural vehicle or deployed on a construction site. The appropriate specification for each of those environments is not the same, and an analysis calibrated to the wrong environment may either over-qualify a product that could be cheaper, or under-qualify one that will fail in the field.
Your question should always be: is this specification derived from measured field data, or is it a standard test level adopted for convenience? Both are legitimate, but they have different implications for how you interpret the analysis results and how much confidence you can place in them.
2.2 The Major Standards Landscape
IEC 60068-2-64: Random Vibration
This is the primary international standard for random vibration testing of electronic equipment. It defines test procedures, severity levels, and documentation requirements. The standard provides guidance on selecting test levels from a set of defined severities, but it is fundamentally a procedure standard — it tells you how to test, not what level to test to. The appropriate severity must be determined from the product's service environment, either by measurement or by reference to documented application profiles.
IEC 60068-2-64 should be read alongside IEC 60068-1, which covers the general principles of environmental testing, and alongside the applicable product standard for your device category. For Zebra-class products, additional application-specific requirements from customer specifications or industry standards may overlay the base IEC requirements.
MIL-STD-810, Method 514: Vibration
Method 514 is the US military standard for vibration testing, and it is widely used in the industrial electronics world because it provides well-documented, physically grounded vibration environments derived from extensive field measurement programs. Procedure I covers general vibration, while Procedure II addresses vehicles and transportation environments. Annex C of Method 514 contains the LCEP — the Life Cycle Environmental Profile methodology — which is the intellectually correct framework for deriving a test specification from the actual service environment.
When a customer invokes MIL-STD-810 for a commercial product, ask whether they have developed a tailored test plan. The standard explicitly states that it should not be applied wholesale — the test levels and durations should be selected to represent the specific service environment. A product that sees MIL-STD-810 cited in a specification without evidence of tailoring is a product whose test level may or may not be appropriate. This is a conversation worth having before the analysis begins.
One nuance specific to Method 514 is the distinction between functional testing and endurance testing. Functional testing evaluates whether the product operates correctly during vibration. Endurance testing evaluates whether it survives a defined lifetime of vibration exposure. These require different analysis approaches — the first is a response analysis at the specified input level; the second is a fatigue life calculation over a defined test duration or service life.
IEC 60721: Classification of Environmental Conditions
IEC 60721 provides a systematic framework for classifying the environments that products encounter throughout their life cycle — from manufacture through transportation, storage, and in-service use. The vibration-related parts of this standard, particularly IEC 60721-3-2 (transportation) and IEC 60721-3-5 (ground vehicle installations), define environmental categories with associated vibration severity levels.
This standard is particularly useful when the customer does not have a specific vibration specification but knows the application category. A product intended for Class 5M2 environments — vehicle-mounted electronics in rough terrain applications — has a defined severity range you can use to frame the analysis even before a formal specification is developed.
ASTM D4169 and ISTA Protocols: Transportation Vibration
When the vibration concern is the shipping and distribution environment rather than in-service use, ASTM D4169 and the ISTA test protocols become the relevant references. These define vibration profiles based on the transportation method — truck, rail, air — and the product weight class. They typically use random vibration profiles derived from measurements of actual vehicle environments.
The key difference between transportation vibration and in-service vibration is duration and severity. Transportation events are relatively short in duration but can be severe in amplitude, particularly for less-controlled environments like courier delivery. In-service vibration for a device mounted in a vehicle accumulates over a much longer period. Your analysis approach, and particularly your fatigue life calculation, must reflect which scenario you are evaluating.
ISO 16750: Road Vehicles — Electrical and Electronic Equipment
For products installed in road vehicles — including vehicle-mounted industrial computers and handhelds used in logistics and fleet management applications — ISO 16750-3 defines the mechanical load testing requirements including vibration. This standard is derived from extensive vehicle measurement programs and is differentiated by mounting location within the vehicle, reflecting the significant variation in vibration severity between locations near the engine, at body mounting points, and at body extremities.
If your customer is specifying a product for in-vehicle use and has not referenced ISO 16750, it is worth asking whether it should be in scope. Products certified to IEC 60068-2-64 levels may or may not meet ISO 16750 requirements depending on the specific severity levels chosen.
2.3 Translating the Specification into Analysis Inputs
Once the applicable standard and severity level are established, every element of the specification must be explicitly mapped to a model input. The PSD table defines the input loading. The frequency range defines the modal analysis scope — you must capture all resonances within and somewhat beyond the input frequency range. The test axes define the set of analyses required. The test duration, combined with assumptions about the natural frequency distribution, defines the number of fatigue cycles to evaluate. The temperature range defines the material property set required.
Write this mapping down before you build anything. It is the engineering bridge between the physical specification and the computational model, and it is the document that allows anyone reading your analysis to understand the physical basis for every modeling choice.
The specification is not just an input file. It is a physical statement about the environment the product must survive. Every modeling choice should be traceable back to something that specification is saying about the real world.
3. Understanding the Ask: What the Customer Usually Means
Random vibration requests come in several flavors, and the flavor matters enormously for how you scope the analysis. The most common request is: can you run a random vibration analysis on this design? That sentence can mean at least four different things, and without clarification, you cannot know which one it is.
It might mean: we need to verify compliance with a customer specification before we can ship. It might mean: we have a field failure that we think is vibration-related and we need to understand why. It might mean: we are in early design and want to understand the dynamic behavior of the structure to guide design decisions. Or it might mean: our customer asked for a vibration simulation in the statement of work and we need to produce one. Each of these requires a fundamentally different analysis, a different level of fidelity, and a different interpretation framework.
3.1 Compliance Qualification vs. Design Guidance vs. Failure Investigation
Compliance qualification analysis is the most common and the most constrained. The specification is known, the product design is essentially fixed, and the question is binary: does the design meet the requirement? This type of analysis requires the highest fidelity because the result will be used to make a gate decision. Material data must be validated. Failure criteria must be clearly defined and physically justified. The model must capture all geometrically and materially significant features of the design. Uncertainty must be quantified and its effect on the conclusion must be assessed.
Design guidance analysis is earlier in the development cycle and the question is softer: where are the dynamic sensitivities, and how can the design be improved? This type of analysis can tolerate somewhat coarser modeling because the goal is relative comparison and trend identification, not absolute prediction. Even so, the modal analysis must be credible — if you miss a natural frequency because your model is too stiff or too coarse, you will draw incorrect conclusions about the design's dynamic behavior.
Failure investigation analysis is reactive. Something has broken and the question is why. This typically requires the most physical intuition of the three, because the model must be built to reproduce the observed failure mode — and if the model does not reproduce it, you have to understand why before you can trust any conclusions. Field failure investigations almost always reveal something the specification or the design analysis did not anticipate, and that lesson is usually more valuable than the formal analysis result.
3.2 The Questions the Customer Cannot Always Answer — But You Must Ask
In vibration work, the following questions reveal the most important information about the analysis scope and the appropriate level of confidence in the results. Some customers will not have answers to all of them. That is useful information too — it tells you where the analysis is most vulnerable to assumption error.
What is the actual service environment, and how was the specification derived? Measured field data, standard table lookup, customer requirement, or inherited from a prior program? Each carries different confidence levels.
What failure mode are you concerned about? Solder joint fatigue, connector fretting, housing cracking, component lead fatigue, PCB fracture, display delamination? Each has a different critical location and a different failure criterion.
What is the product's service life, and how much of that life involves vibration exposure? A device carried by a worker for eight hours per day experiences far more cumulative vibration than one that is occasionally transported. The fatigue life calculation depends critically on the exposure duration.
Has the product, or a predecessor, been physically tested? What happened? Pass? Fail? Marginal? Where did it fail, and at what point in the test? This data is invaluable for calibrating the simulation approach.
What does success look like, and what will you do if the analysis predicts a problem? If the design is already frozen, a negative result means a specification waiver request or a physical test regardless. Understanding this shapes how conservative the analysis needs to be.
Are there known resonances or vibration sensitivities in similar products? Experienced product teams know which features have historically been vibration-sensitive. This prior knowledge should inform where you focus model fidelity.
What is the thermal environment during vibration exposure? Temperature affects damping, stiffness, and fatigue properties for polymers and solder alloys. A room-temperature analysis for a product that vibrates at elevated operating temperatures may be unconservative for certain failure modes.
3.3 The Ask That Does Not Make Sense Yet
In random vibration, the request that is most often premature is a fatigue life prediction on a design that has no physical test correlation, no validated material fatigue data, and no documented failure criterion. The customer wants a number — a predicted life in hours or cycles — and they want it to be definitive.
The honest answer is that a fatigue life prediction for random vibration can easily span two orders of magnitude depending on the S-N curve used, the assumed stress concentration factor, the surface finish correction, and the mean stress correction. A prediction of fifty thousand hours and a prediction of five hundred hours can both emerge from the same stress analysis, depending on which material data and fatigue model assumptions are applied. Presenting either number as a design verdict without quantifying that uncertainty is misleading, regardless of how sophisticated the finite element analysis is.
When this situation arises, the path forward is not to refuse the analysis. It is to reframe it. Instead of a single fatigue life prediction, present the analysis as a sensitivity study: here are the stress amplitudes at the critical locations, here are the dominant frequencies driving the damage, here is what the fatigue life would be under conservative, median, and optimistic material assumptions. This is more information, honestly presented, and it gives the design team what they actually need — an understanding of the physical margin and the key uncertainties — rather than a false sense of certainty.
A fatigue life prediction is only as good as the material data and failure criterion behind it. When those inputs are uncertain, the analysis should communicate that uncertainty explicitly. A range is more honest than a number, and a sensitivity analysis is more useful than false precision.
4. Being a Collaborative Partner in the Vibration Space
Vibration analysis sits at an unusual intersection in the product development process. The physics is grounded in structural dynamics — modal analysis, frequency response, fatigue — but the real-world meaning of the analysis depends on manufacturing quality, material consistency, assembly variation, and a dozen other factors that the structural analysis alone cannot capture. Effective collaboration in vibration work means engaging across those boundaries.
4.1 Connecting Simulation to the Physical Test Program
Random vibration analysis and physical vibration testing are not alternatives. They are partners in the qualification process, and the value of each is greatly amplified when they are planned together.
The simulation tells you where the critical frequencies and critical locations are before any hardware is built. It guides the placement of accelerometers and strain gauges in the physical test. It predicts which design variants will have better dynamic behavior, allowing the test program to focus on the most promising designs. It provides a baseline against which to interpret anomalous test results.
The physical test validates the simulation. It confirms predicted natural frequencies or reveals discrepancies that point to modeling errors or manufacturing differences. It provides the correlation data that is essential for establishing confidence in the simulation approach for future designs. And it sometimes reveals failure modes that the simulation did not predict — which is itself an important learning that improves both the simulation capability and the physical understanding of the product.
If your customer is planning a physical test program, be involved in that planning from the beginning. Help define the accelerometer placement based on the simulation. Help identify the measurement quantities that will best distinguish between competing design hypotheses. Help plan the test sequence so that the most informative measurements are made before fatigue accumulation compromises the specimen.
4.2 Educating the Customer on What the Analysis Can Say
Many of the customers who request random vibration analyses have a fundamentally deterministic mental model of the world. They think of structural analysis as: apply load, get stress, compare to allowable, pass or fail. Random vibration does not work that way, and part of your collaborative role is to help them understand the difference.
The most important concept to communicate is that random vibration analysis is inherently probabilistic. The stress response is described by a distribution, not a single value. The fatigue damage is a statistical expectation, not a certainty. A product that analysis predicts will survive one million cycles has some nonzero probability of failing before one million cycles — the question is whether that probability is acceptable.
This probabilistic framing often makes customers uncomfortable because it does not produce the decisive yes or no they are looking for. But it is the truth of the physics, and helping customers understand it leads to better engineering decisions. It shifts the conversation from does it pass to what is the margin and where is the uncertainty, which are the right questions.
Explaining the significance of natural frequencies to a non-specialist is the most valuable communication you can do in this space. The concept that a structure amplifies vibration at its resonant frequencies — and that a natural frequency landing inside the input frequency band means the structure is being driven at its most sensitive point — is one that most engineers outside the dynamics specialty do not intuitively grasp. Once they do, design conversations change fundamentally. They start thinking about mass and stiffness as tools for shifting resonances, not just about strength.
4.3 The Material Data Gap in Vibration Work
Vibration fatigue analysis is severely limited by the availability of relevant material data, and this is a gap that the simulation community has not solved and often under-acknowledges. The S-N curve data commonly available in handbooks represents idealized specimens, typically smooth round bars, tested in fully reversed bending or tension-compression at room temperature. Real product features — the stress concentrations at connector mounting holes, the solder joint geometry at component pads, the weld lines in molded housings — bear almost no resemblance to those specimens.
When you apply a smooth-bar fatigue curve to a stress concentration in a product, you are implicitly assuming that the Kt factor fully accounts for the notch sensitivity of the material and the surface condition effect. For ductile metals, this is often approximately correct. For brittle materials, heavily filled polymers, and solder alloys, it can be dramatically unconservative.
The collaborative answer to this problem is to work with the design team and the materials engineering function to develop test data that is relevant to the actual product features and materials. Coupon-level fatigue testing of representative geometries — even at modest sample sizes — is more valuable for the simulation than any amount of literature data. If budget or time does not permit dedicated fatigue testing, at minimum acknowledge the data gap explicitly in the analysis and assess its effect on the conclusion.
The S-N curve is the foundation of the fatigue life prediction. If that foundation is uncertain — and for most product-relevant geometries and materials, it is — the predicted life number is uncertain by at least the same factor. Know your data, know its limitations, and communicate both.
5. Before the First Button Is Pushed: The Path Forward
The sequence matters in vibration analysis more than in almost any other simulation type, because the results of each step define the scope of the next. A modal analysis that misses a natural frequency will produce a response analysis that underestimates damage. A response analysis with incorrect damping will produce a fatigue life prediction that is wrong by the square of the damping error in the vicinity of resonance. Getting the foundations right before moving to the next step is not optional.
5.1 The Pre-Analysis Checklist
Work through every item below before committing to the analysis scope. As with the drop test guide, the goal is not to create bureaucratic overhead. It is to force the conversations and decisions that determine whether the analysis will be physically meaningful.
Problem Definition and Context
• The analysis type is defined: compliance qualification, design guidance, or failure investigation.
• The failure mode of concern is physically described — solder joint fatigue, housing crack, connector fretting, PCB flexure fatigue, or other.
• The decision this analysis will support, and the decision gate it feeds, are documented.
• The stakeholders and their requirements for the analysis are identified.
Specification and Standards
• The applicable standard (IEC 60068-2-64, MIL-STD-810 Method 514, ASTM D4169, ISO 16750-3, or customer specification) is identified.
• The PSD table has been reviewed for internal consistency: the area under the PSD integrates to the stated g RMS squared.
• The test axes (X, Y, Z, or combinations) are specified.
• The test duration and its physical basis (service life equivalent, accelerated test factor, or standard table value) are documented.
• The temperature during vibration exposure is specified.
• The specification provenance is understood: measured field data, standard severity table, or inherited requirement.
Modal Analysis Foundation
• The frequency range for the analysis is established: at minimum from the lowest frequency in the PSD to at least twice the highest frequency.
• The model boundary conditions reflect the actual mounting configuration of the product as closely as possible — not idealized fixed or free conditions unless those are physically justified.
• The mesh density is sufficient to accurately capture modes up to the upper end of the analysis frequency range. A rule of thumb: at least six elements per wavelength at the highest frequency of interest.
• The structural damping assumption is documented and physically justified. For electronics assemblies, damping is often in the range of one to three percent critical damping, but this varies with assembly method, connector presence, and thermal interface materials.
• The mass distribution of the model has been verified against the physical product, including circuit boards, connectors, batteries, displays, and other heavy subassemblies.
Material and Fatigue Data
• Material properties are available for all structurally significant components and subassemblies.
• The S-N curve data for the critical material and failure location is identified, and its applicability to the actual product geometry and surface condition is assessed.
• Stress concentration factors for the critical geometric features are documented.
• Temperature effects on fatigue properties are addressed if the operating temperature range requires it.
• The mean stress state at the critical location is assessed and the appropriate mean stress correction (Goodman, Gerber, or other) is identified.
• Any known data gaps are documented and their effect on the fatigue life prediction is bounded.
Acceptance Criteria
• The failure criterion is defined: maximum allowable stress amplitude, minimum fatigue life in hours or cycles, or comparison to a correlated simulation of a design that passed physical testing.
• The physical basis for the criterion is documented.
• The relationship between the criterion and any available physical test data is established.
Correlation and Validation Plan
• Available physical test data for the current or predecessor design has been reviewed.
• Any discrepancies between prior simulation predictions and physical test results are understood.
• The plan for correlating the current simulation to physical test data, if testing is planned, is documented.
5.2 The Modal Analysis Review — Your First Sanity Check
Before computing any vibration response, review the modal analysis results carefully. Look at the mode shapes, not just the frequencies. A mode that shows large motion at a connector, a solder joint array, or a cantilevered component is a mode that will dominate the fatigue damage calculation. A mode that shows rigid-body-like motion of a major subassembly suggests that the mounting interface is too compliant and that the boundary conditions may be questionable.
Compare the predicted natural frequencies to any available test data or to known behavior of similar products. A first mode that seems physically implausible — either far too low, suggesting an overcompliant model, or far too high, suggesting an overstiff model — should be investigated before proceeding. The response analysis and fatigue calculation inherit all of the errors in the modal analysis, and those errors are amplified in the vicinity of resonance.
If natural frequencies fall inside the input PSD frequency range, flag them explicitly in your analysis plan. These modes will dominate the response, and any uncertainty in their frequency or damping translates directly into uncertainty in the fatigue damage prediction. This is also the information the design team needs: a resonance inside the input band is a design risk, and the options for managing it — shifting the frequency through stiffness or mass changes, adding damping, or modifying the mounting interface — are engineering choices that are best made before the design is frozen.
5.3 The Response Analysis and Its Interpretation
Once the modal analysis is validated, the response analysis translates the input PSD through the structure's frequency response to produce response PSDs at every degree of freedom. The RMS stress at each location is extracted from the response PSD and used in the fatigue calculation.
Interpret the response results physically. High RMS stress at a location that has no geometric stress concentration is pointing to a dynamic amplification — a resonance effect. High stress at a location with significant geometry change is pointing to both dynamic and geometric effects. Understanding which contribution dominates is important for identifying the right engineering fix: if the problem is dynamic, shifting the resonance or adding damping will help; if the problem is primarily geometric, changing the local geometry to reduce the stress concentration is more effective.
Present response results with the frequency content intact where possible. A single RMS number tells the design team how much stress there is, but not where in the frequency spectrum the damage is coming from. A response PSD at the critical location, or a plot of cumulative damage versus frequency, tells the story in a way that enables better engineering decisions.
5.4 Communicating Results Honestly
The fatigue life prediction is the number the customer will focus on, and it deserves more context than a single value. Present the sensitivity of the prediction to the key assumptions: damping ratio, stress concentration factor, and S-N curve choice. If a factor of two change in the damping assumption changes the predicted life by an order of magnitude — which it will if there is a resonance inside the input band — the customer needs to know that.
If the analysis predicts that the design passes with significant margin, say so clearly — but also identify the assumptions under which the margin disappears. If the analysis predicts a marginal result, do not round it to a pass. A marginal prediction is a signal that the design is sensitive and that the physical test result may be different from the simulation result in either direction.
If the analysis predicts a failure, present it as the beginning of a design conversation, not the end of an analysis engagement. Where is the damage concentrated? What is driving it — the resonance frequency, the damping level, the geometry? What design changes would most effectively reduce the damage? This is where the simulation earns its value: not by pronouncing judgment, but by diagnosing the problem clearly enough that the design team can fix it.
The most useful random vibration analysis is one that tells the design team where the energy is going and why. A pass or fail verdict without physical diagnosis is a missed opportunity, regardless of what the verdict is.
6. A Final Word on Physical Intuition
Random vibration analysis is one of the disciplines where the gap between analysts who have physical intuition and those who do not is most consequential. The mathematics can be executed correctly — the PSD can be integrated, the frequency response functions can be computed, the fatigue damage can be summed — and the result can still be physically wrong if the analyst does not have a felt sense of what the numbers mean.
Physical intuition in vibration work is built through experience with real products and real test data. It comes from watching a physical test and being able to predict which component will fail before the test is finished, based on what the modal analysis revealed. It comes from seeing simulation predictions that disagreed with test results and understanding why — what was the model missing, what did the test reveal that changed the physical picture. It comes from handling the hardware and recognizing the features that will be problematic from a dynamic standpoint before any analysis is run.
If you are early in your career in this discipline, seek out that experience actively. Attend physical tests when you can. Talk to test engineers about what they observe. Ask to see the strain gauge data from historical tests. Study the failure modes in actual field returns. Every one of those experiences is adding to the physical model you carry in your head, and that model is what separates analysis from computation.
The software will continue to improve. The numerical methods will become more automated and more capable. The boundary between what requires human judgment and what can be computed will shift. But the judgment about whether the physics of the model reflects the physics of the product — and whether the conclusion of the analysis serves the engineering decision it is meant to support — will always require a person who understands both.
Build that understanding deliberately. Question every assumption. Validate every prediction you can. And remember that the purpose of all of this is not to produce a simulation. It is to build products that work in the field, for the people who depend on them.
A simulation that confirms what you already believe is not engineering. A simulation that tells you something you did not know — about where the resonances are, what the critical failure mode is, where the design margin actually lives — is worth every hour of careful, thoughtful work that produced it.
Appendix A: Standards Quick Reference
IEC 60068-2-64 (Random Vibration): Primary international standard for random vibration testing of electronic equipment. Defines test procedures and guidance for severity selection. Must be paired with field environment data or application-specific standards for appropriate severity level selection.
MIL-STD-810, Method 514 (Vibration): US military environmental standard with extensive vehicle and transportation vibration profiles derived from field measurement. Procedure I for general vibration, Procedure II for vehicle environments. Emphasizes tailoring to actual service environment through the LCEP methodology.
IEC 60721-3 (Classification of Environmental Conditions): Systematic framework for environmental severity classification by application category. Parts -3-2 (transportation) and -3-5 (ground vehicle installation) are most relevant for handheld and vehicle-mounted electronics.
ASTM D4169 (Shipping Container Performance): Random vibration profiles for transportation environments, differentiated by transportation mode and product weight class. Used when evaluating distribution cycle vibration rather than in-service vibration.
ISTA Test Protocols: Complementary to ASTM D4169, developed by the International Safe Transit Association. Widely used in the packaging and distribution community. Protocols 2A, 2B, and 3A are commonly applied to packaged product vibration testing.
ISO 16750-3 (Road Vehicles, Mechanical Loads): Vibration requirements for electrical and electronic equipment installed in road vehicles. Differentiated by mounting location. Relevant for any product intended for in-vehicle use.
IEC 60068-1 (General Principles of Environmental Testing): Foundational standard covering the principles and methodology underlying all IEC 60068 series tests. Should be read before applying any specific test method.
Appendix B: Pre-Analysis Conversation Template
Use the following questions to guide the first conversation with any random vibration customer. Not all questions apply to every situation. The goal is shared physical understanding, not interrogation.
• What is the product, and what service environment will it experience vibration in?
• What failure mode are you concerned about — fatigue fracture, connector wear, solder joint failure, display delamination, other?
• What standard or specification defines the test requirement, and where did that specification come from?
• What are the test axes, the frequency range, and the overall g RMS level?
• What is the test duration, and what service life equivalent does it represent?
• What temperature does the product operate at during vibration exposure?
• Does the product have any known natural frequencies from previous testing or analysis?
• Has the product, or a similar design, been physically tested for vibration? What was the outcome?
• Are there field failure reports that suggest vibration-related damage?
• What material fatigue data is available for the critical components and materials?
• What does a passing result look like, and where does that acceptance criterion come from?
• What decisions will this analysis support, and when do you need results?
Joseph P. McFadden Sr. | Engineering Fellow, MEAS | Zebra Technologies
McFaddenCAE.com | Building Intuition Before Equations
