The Gold Standard of Vibrations | Why an Unknown Driving Force Prevents Damping Identification

The output in our examples is the displacement x(t) of one modelled mass relative to a fixed equilibrium reference. The input is the external force F(t). They are related through the mass, stiffness and damping, but they are not the same variable.

A graph of x against time can show amplitude, timing and changes in the observed motion. It does not directly display the applied force. A large movement might occur under a large input, a strongly amplified response, or a combination of both.

Draw a box with force entering and displacement leaving. Put m, c and k inside the box as model parameters. That picture prevents a common interpretive jump: treating the output curve as though it already contained a separately measured input.

The question “What force would make this model follow the supplied motion?” is a forward balance calculation once the parameters are specified. The question “Which parameters produced this motion under an unknown force?” is an inverse problem. The second has missing information that the first calculation can help expose.

Contents · Next chapter

Our equation is m x″ + c x′ + kx = F(t). Mass m is positive, stiffness k is positive and viscous damping coefficient c is nonnegative. The model has one degree of freedom and linear, constant coefficients.

Displacement is in metres, time in seconds, force in newtons, mass in kilograms, stiffness in newtons per metre and c in newton-seconds per metre. Each term on the left has units of force. This unit check helps distinguish damping coefficient from a dimensionless damping ratio.

The equation represents a chosen idealisation. Real systems may have several modes, nonlinear restoring forces, changing boundaries or other dissipation mechanisms. An identification claim must first establish why the chosen model is adequate for the question.

The OpenStax forced-oscillation chapter provides the introductory physical setting. We will not assume that its familiar sinusoidal forcing is known in an arbitrary measured record. That is precisely the information whose absence matters here.

Contents · Previous chapter · Next chapter

Let the supplied motion be x(t) = 0.010 cos(2t) metres, with t in seconds. Its amplitude is ten millimetres and angular frequency two radians per second. Its period is π seconds.

Differentiating gives x′(t) = −0.020 sin(2t) metres per second and x″(t) = −0.040 cos(2t) metres per second squared. These are exact derivatives of the stipulated analytic motion, rather than estimates from noisy samples.

At t = 0, displacement is 0.010 m, velocity zero and acceleration −0.040 m/s². At t = π/4, displacement is zero, velocity −0.020 m/s and acceleration zero. These two states will later provide different information about stiffness and damping.

Before inserting the motion into a force balance, keep the unit conversions visible. Ten millimetres is 0.010 metres, not ten metres. A thousandfold displacement mistake would propagate into every reconstructed force and energy. The model’s neat appearance does not protect it from that basic representation error.

Contents · Previous chapter · Next chapter

Choose m = 1 kg, k = 4 N/m and c = 0. Substituting the supplied motion gives m x″ + kx = −0.040 cos(2t) + 0.040 cos(2t), which is zero.

Thus F(t) = 0 is compatible with this parameter set and this motion. It is an undamped free oscillator with the matching initial displacement and velocity. Its natural angular frequency is the square root of k/m, equal to two radians per second.

This is one valid explanation under the model, not yet a unique explanation of the trace. The calculation establishes compatibility. To establish uniqueness, we would need to show that every other allowed parameter-and-input combination fails.

That distinction is worth practising. A successful substitution verifies a candidate; it does not prove that the candidate is the only possibility. The next chapter retains the exact same motion while changing the damping and input, supplying a direct counterexample to uniqueness.

Contents · Previous chapter · Next chapter

Keep m = 1 kg and k = 4 N/m, but set c = 1 N s/m. The mass and spring terms still cancel. The remaining force is c x′ = −0.020 sin(2t) newtons.

If that external force is supplied, the damped model follows the same displacement x(t) = 0.010 cos(2t). Its output amplitude and frequency match the undamped explanation exactly. Its physical energy account differs because the force replaces the energy dissipated by damping.

The applied force here has the same sign as velocity. It does positive work over the cycle. The damping force acting on the mass is −c x′, opposite to velocity; moving the damping term to the left side of the equation must not obscure that sign.

We have therefore constructed two complete model explanations: zero damping with zero external force, and positive damping with a matched external force. Observing the displacement alone cannot distinguish them if both force histories are allowed. More decimal places in the same exact output trace would not resolve this structural ambiguity.

Contents · Previous chapter · Next chapter

Choose a third explanation with m = 1 kg, k = 9 N/m and c = 1 N s/m. Substitution now gives F(t) = 0.050 cos(2t) − 0.020 sin(2t) newtons.

The stiffness term exceeds the mass term by 0.050 cos(2t). Adding the damping term supplies the remaining sine contribution. With this force, the third system again follows the exact same displacement.

The table compares compatible complete explanations, not three parameter estimates derived independently from one output. It shows which extra input changed when a parameter changed. Keeping that column visible is essential to understanding the ambiguity.

Contents · Previous chapter · Next chapter

For a supplied sinusoidal motion x = A cos(ωt), substitution gives F(t) = A(k − mω²)cos(ωt) − cAω sin(ωt). Any allowed choice of the coefficients defines a force history compatible with that motion.

This does not mean every such force history is physically available in a real experiment. It means that if the input is left unconstrained, the displacement equation alone cannot exclude those mathematical alternatives. Independent input restrictions can narrow the allowed set.

The same construction works beyond a sinusoid. For any sufficiently differentiable supplied x(t), choose model coefficients and calculate F(t) from m x″ + c x′ + kx. A known output plus unconstrained input generally leaves a large family of parameter-and-force pairs.

A good inverse-problem statement therefore lists what is unknown on both sides. Writing only “find c” can hide the fact that an entire function F(t) is also unknown. The number of recorded displacement points does not by itself remove that freedom.

Contents · Previous chapter · Next chapter

For the oscillator, mechanical energy is E = ½m(x′)² + ½kx². Differentiating and using the equation of motion gives dE/dt = F x′ − c(x′)². External power enters through F x′ and dissipation through c(x′)².

For Explanation B, F = c x′. The two power terms cancel at every instant, so mechanical energy remains constant despite positive damping. With k = 4 and amplitude 0.010, the constant energy is 0.0002 joules.

The mean dissipated power over a full sinusoidal cycle is cA²ω²/2. Our values give 0.0002 watts. Multiplying by period π gives about 0.000628319 joules dissipated per cycle and the same amount supplied by the force.

This energy check independently supports the force-balance construction. It also corrects the shortcut “constant amplitude proves no dissipation.” Constant amplitude can instead reflect a balance between supply and loss. To decide which explanation applies, the input or energy transfer needs additional evidence.

Contents · Previous chapter · Next chapter

In Explanation B, the external force varies sinusoidally and is zero at the displacement extremes. At those instants velocity is zero, so both instantaneous input power and damping dissipation are zero.

Near equilibrium crossings, velocity magnitude is greatest. The required damping-compensating force and instantaneous power are correspondingly largest. The force is coordinated with the motion; it is not a constant push in one direction.

A nonzero force does not necessarily do work at an instant when velocity is zero. Conversely, a small force acting in the direction of velocity over time can supply a positive amount of energy. Force, power and accumulated work need their separate definitions.

A useful learning task asks for the signs of force, velocity and power at several points in the cycle before calculating an integral. That connects the phase relationship with the energy account. It also makes it harder to confuse a damping coefficient with the height of a displacement graph.

Contents · Previous chapter · Next chapter

A shrinking envelope does not automatically identify damping if forcing is unknown. Consider x(t) = 0.010 exp(−0.5t)cos(2t) metres. This motion is compatible with a free model having m = 1, c = 1 and k = 4.25 in the corresponding SI units.

The exponent’s decay rate is 0.5 per second, and the damped angular frequency is two radians per second. Their relationship to the free linear equation can be checked by differentiating and collecting the cosine and sine terms.

Now change c while retaining the same x. An additional force equal to the change in c times x′ restores the balance. Changing k adds another force contribution equal to the change in k times x. The same decaying trace can therefore remain compatible with other coefficients under changed inputs.

This does not invalidate free-decay analysis. It identifies a premise that such analysis needs: the relevant external forcing is absent or adequately accounted for over the fitted interval. That premise comes from the experiment and model, rather than from the decaying curve alone.

Contents · Previous chapter · Next chapter

A free response means that the model’s external force is zero over the interval in question. Initial energy can still produce motion. “Free” does not mean zero displacement, zero velocity or zero internal forces.

If independent evidence supports F = 0 and the linear single-mode model is suitable, the motion can constrain the coefficients more strongly. A damping estimate derived under that condition should retain the condition in its interpretation.

A command to stop an actuator is not necessarily identical to proof that every relevant force has vanished. Nor does a quiet-looking environment establish an ideal isolated oscillator. A real investigation needs the appropriate input, coupling and boundary evidence.

For a classroom problem, simply state the zero-force condition as part of the question. Then later remove it and ask which conclusion becomes unresolved. This controlled change teaches the role of an assumption more effectively than repeating a decay formula with new numbers.

Contents · Previous chapter · Next chapter

Return to the sinusoidal motion with known mass one kilogram. Suppose force is independently supplied as F(t) = −0.020 sin(2t) N and the linear constant-coefficient model is stipulated. Rearrange the equation as c x′ + kx = F − m x″.

At t = 0, velocity is zero and displacement 0.010. The right side is 0 − (−0.040) = 0.040 N. Thus k × 0.010 = 0.040 and k = 4 N/m.

At t = π/4, displacement and acceleration are zero while velocity is −0.020. The force is −0.020 N, so c × (−0.020) = −0.020 and c = 1 N s/m.

The added input information distinguishes Explanation B from A and C within this model. We did not infer force from the desired parameter and then call it independent evidence. It was supplied separately, making the comparison a genuine change in the available information.

Contents · Previous chapter · Next chapter

Two records are not automatically two independent constraints. In the rearranged equation c v + kx = y, the rows contain velocity and displacement. To separate c and k with known mass and force, the rows must provide sufficiently different combinations.

For two exact states, the determinant v₁x₂ − v₂x₁ must be nonzero. If both states are equilibrium crossings, both have x = 0 and cannot determine k from those rows alone. If both are turning points with v = 0, they cannot determine c.

Our selected states deliberately isolate the coefficients. One has x nonzero and v zero; the other has x zero and v nonzero. That makes the arithmetic transparent and the information requirement visible.

In noisy records, a determinant that is merely nonzero but extremely small can still produce an unstable estimate. This is a practical conditioning issue distinct from exact nonuniqueness. Identify which problem is present before promising that more calculation will solve it.

Contents · Previous chapter · Next chapter

The known-force example assumed mass one kilogram. If mass is also unknown, the equation contains another coefficient multiplying acceleration. A single-frequency sinusoid has x″ = −ω²x, so mass and stiffness appear together as k − mω².

Even with that sinusoidal force record, the cosine component alone cannot separate m and k at one frequency. Different positive pairs with the same difference can remain compatible. The sine component can constrain c when amplitude and phase are known, but it does not automatically resolve the mass–stiffness ambiguity.

Additional mass evidence or suitably different motion conditions can supply new constraints. For example, records at different known frequencies can distinguish combinations that one frequency merges, provided the model parameters remain the same and the input records are supported.

The lesson is to revisit the unknown list whenever a supplied constant becomes uncertain. A solution that was unique with known m does not remain unique merely because the same algebraic symbols are printed on the page.

Contents · Previous chapter · Next chapter

A sinusoidal force and displacement have amplitudes and a relative phase. Retaining that phase provides information about the balance of stiffness, inertia and damping. Reporting only the displacement amplitude discards part of the relation.

For our Explanation B, the force is proportional to velocity and is a quarter-cycle out of phase with displacement under the chosen convention. That timing is what allows it to replace damping losses throughout the cycle.

If a report gives only a force amplitude and a displacement amplitude, several phase relationships may remain allowed. Their ratio is not by itself the full complex frequency response. The force and displacement must also refer to the same frequency, location model and timing reference.

A paper exercise can supply two force traces of equal amplitude with different phases. Ask which is compatible with the stated displacement and coefficients. This tests whether the learner can use the complete relationship rather than selecting a model from the tallest curve.

Contents · Previous chapter · Next chapter

Our exact derivatives came from an analytic teaching function. Real displacement samples do not arrive with exact velocity and acceleration values attached. Estimating derivatives can magnify noise and depends on sampling, filtering and the fitted representation.

A smooth fitted curve can make differentiation possible, but the derivative then inherits the fit’s assumptions. It is not a new independent measurement simply because it occupies another column. Displacement, derived velocity and derived acceleration may share errors.

The Vibrations root explains sampling and spectra. Use that route when the main gap concerns how acquisition can misrepresent motion. Here we assumed exact continuous functions to isolate a different limit: even perfect output information can leave coefficients unresolved when input is unconstrained.

Keeping those two limitations separate is useful. Better sampling addresses a measurement problem. Independent force evidence addresses an input ambiguity. A well-designed investigation identifies which missing information its next measurement can actually provide.

Contents · Previous chapter · Next chapter

Exact ambiguity means different allowed explanations give the same output under the model. Ordinary measurement uncertainty means the output itself is not known exactly. Both can occur together, but they require different remedies.

Our A, B and C explanations match x(t) at every time, not merely within an error bar. Repeating the same output measurement with a smaller error bar therefore cannot separate them while their corresponding forces remain allowed.

If force were independently bounded, some explanations could be excluded. For instance, a justified bound on available force amplitude below the requirement of C would rule C out. It might still leave both A and B, depending on the bound and other conditions.

A result should state the remaining candidate set rather than forcing one preferred value. “The record supports these combinations under these input constraints” is more informative than a highly precise coefficient whose apparent uniqueness came from silently fixing an unknown force.

Contents · Previous chapter · Next chapter

Given chosen coefficients and known force, calculate the residual r(t) = m x″ + c x′ + kx − F. A zero residual verifies compatibility with the exact model record. A nonzero residual reveals disagreement requiring investigation.

A small residual in measured data does not prove that the model is physically complete. Flexible inputs, overfitting, filtering or an inappropriate error model can also produce close agreement. Validation requires attention to independent evidence and the intended prediction question.

If F is freely reconstructed from the same x and chosen coefficients, the residual is zero by construction. Treating that zero as evidence that the coefficients were discovered from the data would be circular. The force calculation demonstrates a compatible explanation, not independent validation.

Preserve what was measured, what was assumed and what was reconstructed. Those labels make a residual useful. Without them, an impressive fit statistic can conceal the very freedom that prevents the intended parameter inference.

Contents · Previous chapter · Next chapter

Use x(t) = 0.020 cos(3t) metres and known mass two kilograms. Suppose the independently supplied force is F(t) = 0.240 cos(3t) − 0.180 sin(3t) newtons. Find the constant c and k in the stipulated model.

Velocity is −0.060 sin(3t) and acceleration −0.180 cos(3t). The inertial force is −0.360 cos(3t). Matching sine terms gives −0.060c = −0.180, hence c = 3 N s/m.

Matching cosine terms gives 0.020k − 0.360 = 0.240, hence k = 30 N/m. Substitute both into the full equation to check every term. The natural angular frequency would be √15, about 3.873 radians per second, which differs from the supplied driving frequency three.

Now remove the force record. The same displacement no longer uniquely fixes those coefficients because another force can be reconstructed for another allowed parameter set. Explaining what changed in the evidence is part of the transfer task.

Contents · Previous chapter · Next chapter

Suppose someone proposes measuring the same displacement for twice as long. In our ideal exact construction, all three explanations predict the same continuing motion. Extending the record does not distinguish them unless it introduces new input or model restrictions.

Suppose instead the proposed evidence independently characterises the applied force, or supplies a supported interval during which force is zero. Those observations can change the admissible equations and remove some alternatives. Their usefulness follows from differing candidate predictions.

The practical question is not “Will we have more data?” but “Which currently compatible explanations will this evidence exclude?” A longer record can be valuable, but only for questions the record is capable of answering.

For learning, make a two-column prediction table before choosing an additional observation. If every candidate gives the same prediction, the proposed check cannot discriminate among them by itself. This habit transfers to many scientific inverse problems beyond vibration analysis.

Contents · Previous chapter · Next chapter

A complete teaching record includes the displacement definition, time unit, equilibrium reference, mass evidence, model equation, coefficient assumptions and input information. Store derived velocity and acceleration with their derivation rather than presenting them as unrelated observations.

In an application record, the measurement locations, sensor direction, acquisition conditions and processing also matter. The connection between the measured point and the single-degree-of-freedom model needs its own justification.

Use Measurement for the broader evidence framework, and Vibration Engineering when the question concerns sources, transmission paths and affected locations. Neither turns this mathematical example into an equipment diagnosis.

A useful final sentence names both the result and its dependency: “These coefficients are compatible with this motion under this independently supplied force and this model.” That wording preserves what the calculation established while leaving room for a later model or measurement correction.

Contents · Previous chapter · Next chapter

For our original motion and known mass one kilogram, the required force has cosine coefficient 0.010(k − 4) and sine coefficient −0.020c. Its peak amplitude is 0.010 times the square root of (k − 4)² + 4c², with coefficients interpreted in the consistent SI units.

Suppose independent information limits that amplitude to at most 0.020 N. Squaring the inequality gives (k − 4)² + 4c² ≤ 4. Together with positive stiffness and nonnegative damping, this is a supported parameter region rather than a single point.

In particular, c cannot exceed one N s/m, and k lies between two and six N/m. Not every combination inside that rectangular range is allowed: the curved inequality couples them. At c = one, k must equal four; at c = zero, the full stiffness interval remains possible.

This example shows why reporting separate parameter intervals can lose dependence. The force bound adds real information and excludes Explanation C, but does not uniquely identify Explanation B. Keep the coupled region when the next decision depends on combinations of stiffness and damping rather than either quantity alone.

Contents · Previous chapter · Next chapter

One sinusoidal record with unknown mass merges stiffness and inertia into k − mω². A second supported record at another frequency can separate that combination if the same constant parameters apply.

Consider two invented motions, each with amplitude 0.010 m, at angular frequencies two and three. Independently supplied force records have cosine coefficients 0.050 N and zero respectively. Dividing by amplitude gives k − 4m = five and k − 9m = zero in consistent units.

Subtracting the equations gives 5m = five, so m is one kilogram and k nine N/m. If the sine coefficients are −0.020 and −0.030 N, each gives c = one N s/m through −cAω. The two damping calculations provide a consistency check under the shared model.

The extra record helps because it changes the coefficient combination, not simply because it adds more rows. If the system changed stiffness between records, the shared-parameter subtraction would no longer be justified. State both the changed frequency and the unchanged model properties before claiming that the new evidence resolves the original ambiguity.

Contents · Previous chapter · Next chapter

The coefficient c has units. A commonly used dimensionless damping ratio for this model is ζ = c/(2√mk). Changing mass or stiffness changes ζ even when c stays fixed.

Explanation B has m = one, k = four and c = one, giving ζ = one-quarter. Explanation C has m = one, k = nine and c = one, giving ζ = one-sixth. Their supplied displacement traces still agree because their forces differ.

A statement that “the damping is the same” therefore needs a named quantity. Equal c does not guarantee equal ζ, equal energy loss under arbitrary motions or equal response curves under a common input. Conversely, an equal dimensionless ratio does not establish equal physical coefficients in different systems.

For a useful check, keep c fixed and multiply both m and k by four. The denominator in ζ becomes four times as large, so ζ becomes one-quarter of its previous value. This is a scaling exercise within the model, not a proposal to alter equipment. Naming the quantity before comparing values prevents a reporting convention from becoming an unsupported physical conclusion.

Contents · Previous chapter · Next chapter

Over a complete steady sinusoidal cycle, the net change of spring and kinetic energy is zero. Therefore the mean external input power equals cA²ω²/2 for this model. This gives another way to constrain c if input-power evidence is independently available.

For the original A = 0.010 and ω = two, mean input power is 0.0002c watts when c is expressed in N s/m. If supported mean power is at most 0.0001 W, c must be at most 0.5 N s/m. Explanation B would be excluded by that bound even though it matches the displacement.

The conclusion depends on a complete-cycle steady account and on the power evidence describing the relevant mechanical input. An electrical supply reading is not automatically identical to mechanical power delivered to the modelled degree of freedom. Conversion losses and other paths need their own account.

This constraint does not independently determine k. Over a complete cycle, conservative storage terms return to their starting values. The example is useful precisely because it shows selective information gain: one additional measurement can constrain damping while leaving another parameter unresolved.

A force bound must also state its amplitude convention. For a pure sinusoidal force, its full-cycle RMS value is its peak value divided by the square root of two. A bound supplied as RMS cannot be substituted unchanged into the peak-amplitude inequality. Convert the convention first, then preserve the coupled parameter constraint. Otherwise a reporting difference can wrongly exclude an otherwise compatible model.

Contents · Previous chapter · Next chapter

If the learner confuses input and output, return to the box diagram and label which quantity each instrument or supplied record describes. If force signs are wrong, draw the damping force opposite to velocity and rebuild the balance.

To stabilise the method, reconstruct forces for several chosen parameter sets using the same motion. Then provide a known force and ask which sets remain compatible. The learner should see why compatibility and identification are different tasks.

For extension, make mass unknown or remove phase information. Ask which parameter combinations the remaining record supports. Another useful task compares two proposed measurements and predicts which one would distinguish the candidate explanations.

Mastery is visible when a student can derive the needed relationship without the original example’s cues, verify a new numerical case and explain a genuine limit. Memorising “damping makes amplitude smaller” is a starting idea. Understanding how input, response and evidence interact makes that idea usable in a changed setting.

Contents · Previous chapter · Next chapter

Does steady amplitude prove zero damping?

No. A suitable input can replace dissipated energy. Our damped and undamped explanations share exactly the same steady displacement.

Does a decaying envelope identify the damping coefficient?

Only with sufficient model and input evidence. A free-decay estimate depends on the relevant force being absent or accounted for, and on the chosen dynamic model being adequate.

Why do two states sometimes fail to identify two coefficients?

They may supply dependent equations. Turning points alone do not reveal the coefficient multiplying velocity; equilibrium crossings alone do not reveal the coefficient multiplying displacement.

Does zero residual prove the chosen parameters are true?

It establishes compatibility with the stated record and model. If the input was freely reconstructed from that same record, zero residual occurs by construction and is not independent identification evidence.

Contents · Previous chapter · Next chapter

Return to Vibrations for motion, damping, forcing and sampling foundations. Use Measurement for interpreting results and Vibration Engineering for the wider source–path–response setting. Actual machine-condition interpretation belongs in the existing Industrial Machinery Vibration Analysis guide.

The OpenStax forced-oscillation treatment provides introductory physics context. The compatible parameter sets, reconstructed forces, energy checks and identification tasks above are original fictional teaching constructions.

The Gold Standard of Education connects clear explanation with independent transfer. The How X Works hub offers other subjects where a measured output leaves more than one possible cause. Preserve the habit of writing the full input–model–output relationship before claiming that one curve has identified its hidden mechanism.

Contents · Previous chapter · Return to Vibrations