Tools › Structures › Loads and Dynamics › Force Limited Vibration › Method

Force Limited Vibration: Method

Open the Force Limited Vibration tool

Method: Semi-Empirical and Two-Degree-of-Freedom Force Limits, Notch Depth, Assumptions

The problem

A conventional vibration test drives the base of the test item to an acceleration specification that envelopes the flight environment. In flight the mounting structure gives way at the resonances of the item: the item acts as a vibration absorber and the interface acceleration dips where the item resonates. A shaker has a very high mechanical impedance, so it holds the envelope through those resonances and overtests the item. Force limiting repairs this by a second specification: the total force between the item and the fixture is measured with force gages, and wherever it would exceed a force limit the controller notches the input. The method is set out in NASA-HDBK-7004C and in its monograph, NASA RP-1403. This tool computes the force limit by the semi-empirical method and the two two-degree-of-freedom methods of those documents and estimates the notch that results.

Sources and revision

The tool follows NASA-HDBK-7004C (30 November 2012); page numbers below are of that revision. Two things are taken from elsewhere because revision C does not print them. The equations of the complex two-degree-of-freedom model are those of RP-1403 (revision C prints only the resulting tables and refers to the journal paper). The default case and the rule that builds residual and modal masses from asymptotes come from the spreadsheet example in Appendix C of revision B (January 2003), which revision C dropped. Revision B's advice to hold the roll-off of a force specification to about 9 dB/octave, its statement that the flight data are enveloped with C about √2, and its recommendation of residual masses for the simple model are not in revision C and are not applied, except that the residual-mass choice remains available as an option.

Units

Frequencies are in Hz and the acceleration specification in g²/Hz (random) or g (sine); these are plain numbers and the unit switch leaves them alone. Masses go through the unit system (lbm or kg). Forces are reported in pounds-force (the weight in lb times g) or newtons (mass × 9.80665 m/s² × g), force spectral densities in lb²/Hz or N²/Hz.

Acceleration specification

Breakpoints are joined by straight lines on log-log axes. A line of the form 2000 -6 dB/oct places that breakpoint on a slope from the previous one (10 log10 per octave for a spectral density, 20 log10 for a sine amplitude). The area of each segment is exact, so the overall g rms and the overall rms force of the semi-empirical limit carry no integration error.

Semi-empirical limit

Handbook section 5.3 (p. 29). For a sine test, Eqs. 1a and 1b: F = C M0 A for f below fb and F = C M0 (fb/f)n A at and above it. For a random test, Eqs. 2a and 2b: SFF = C² M0² SAA and SFF = C² M0² (fb/f)2n SAA. M0 is the total mass of the test item and C a dimensionless constant of the configuration; it takes the place that the amplification Q holds for the item alone on the shaker, so the limit clips the force spectrum at C² instead of Q². The break frequency fb and the exponent n account for the fall of the asymptotic mass of the item with frequency. Before the test fb is usually the fundamental resonance of the item on the shaker, which is how the tool uses the entered f0; n is often equal to unity and seldom less. Both are to be verified in a low-level run without force limiting: fb is the frequency at which the asymptote of the measured force spectrum starts to roll off, and the roll-off should follow that of the asymptotic mass without giving up the broadband high-frequency response. The limit keeps the breakpoints of the specification with fb added.

Guideline 2 of the handbook (p. 40): the semi-empirical method is a framework for comparing flight data and other predictions. It is not to be used by itself with a C² that is arbitrary, based on the mass of the item alone, or chosen to reach a wanted response; the choice needs a rationale from the flight mounting configuration, such as the two models below, a finite-element analysis, or values measured for similar configurations.

Simple two-degree-of-freedom model

Handbook section 5.4 and Appendix B (pp. 31–33, 57): one oscillator for the source and one for the load, tuned to the same uncoupled frequency ω0, which gives the largest interface force. With μ = m2/m1 the coupled resonances are those of the classical dynamic absorber, (ω/ω0)² = 1 + μ/2 ± √(μ + μ²/4), Eq. B1. The ratio of the interface force and acceleration spectra there is the squared apparent mass of the load, SFF/(SAA m2²) = [1 + β²/Q2²]/[(1 − β²)² + β²/Q2²] with β = ω/ω0, Eq. B2, evaluated at the lower resonance where it is slightly larger. This is the curve of the handbook's Figure 8. For a very light load it tends to Q² + 1 (no force limiting); for equal masses it is 2.62; for a heavy load it tends to 1.

The model has one mass per side, so there is a choice of what to enter. Revision C recommends the asymptotic apparent masses of the load and the source, "because the asymptotic apparent mass envelops both the modal and residual masses", and notes that the ordinate of Figure 8 is then the C² of the semi-empirical method. The tool has no measured apparent mass to smooth, so it estimates the asymptotic mass of a band as the residual mass plus the modal mass of that band (which, with masses built from asymptotes, is the residual mass of the band below). For the rod of the handbook's Figure 6 this estimate equals the plotted asymptotic mass at the first mode and lies 15 to 30 percent below it at the third and fifth. This estimate is the tool's own. The other option uses the residual masses alone, as revision B recommended and its example does.

Complex two-degree-of-freedom model

Handbook section 5.5 and Appendix C, Tables 2 to 4 (pp. 33, 58–61); equations from RP-1403 section 4.1.2. The source and the load each keep a residual mass (M1, M2, joined at the interface) and a modal mass on a spring and damper (m1, m2). With μ = M2/M1, α1 = m1/M1, α2 = m2/M2 and the tuning Ω = ω2/ω1, the undamped coupled resonances follow from the quadratic of RP-1403 Eq. 23, β2² = −B/2 ± √(B² − 4C)/2 with B = −[(1 + μ + α1)/Ω² + (1 + μ + μα2)]/(1 + μ) and C = (1 + μ + α1 + μα2)/[(1 + μ)Ω²]. The squared apparent mass of the load is Eq. 24, |M2|²/M2² = {[(1 − β2²) + α2]² + β2²(1 + α2)²/Q2²}/[(1 − β2²)² + β2²/Q2²]. The force peak and the acceleration peak need not fall at the same resonance, so the interface acceleration is computed at both resonances for a flat external force on the source modal mass (Eqs. 25 and 26, with the complex apparent masses M = Mres + m(1 + iβ/Q)/(1 − β² + iβ/Q) that the monograph leaves implicit); the apparent mass at the resonance with the smaller acceleration peak is scaled by the ratio of the peaks, and the larger of the two values is kept. The search is repeated for Ω² from 8/16 to 32/16 in steps of 1/16, as in the monograph, and the largest value is the normalized force limit. It equals 1 when the load has no modal mass in the band. When α1 is below 0.25 the value at 0.25 is used, as the monograph suggests. Both oscillators take the same Q, as in the printed tables; the handbook allows other Q by interpolation or from the equations, and the tool computes from the equations.

Revision C says that where both the residual and the modal masses are known the complex model should be used, because its masses are precisely defined. Revision B recommended computing both models and taking the larger in each band; the tool shows both and marks the larger, and bases the notch estimate on it.

Effective masses by band

The residual mass in a band is the sum of the effective masses of all modes above it; the modal mass is the effective mass of the modes in the band (handbook section 5.2 and Appendix A). They come from a finite-element effective-mass table, or from a smoothed measured apparent mass (shaker run for the item, tap test for the mounting structure). The tool takes them either from a typed table or from asymptotes built as the spreadsheet of revision B's Appendix C builds them: the residual mass equals the total up to a break frequency and falls as (fr/f)p above it, and the modal mass of a band is the drop of the residual mass from the band before. Bands are the nominal one-third-octave centers inside the specification (or the rows of the typed table). Overall rms forces of the band results integrate straight log-log lines through the band values. The handbook notes that the masses usually need to be evaluated only at the fundamental mode of the item on the shaker, and that the force limit should be specified as a smooth, slowly varying curve.

Notch depth and the guidelines

At the resonance of the item on the shaker, a modal mass εM0 on its spring and the remaining (1 − ε)M0 rigid give, from RP-1403 Eq. 24 with β = 1, |M|² = M0²(1 + ε²Q²). The unnotched force is this apparent mass times the specification, and the notch is the ratio of the unnotched force to the limit: A² on a spectral density, A on an amplitude, 20 log A in decibels. The resonance left after the notch is the resonance height less the notch. The mean-square response left by a notch of depth A² at a lightly damped resonance is 1 − (2/π)[tan⁻¹√(A² − 1) − √(A² − 1)/A²], the approximate form printed on the handbook's Figure 4 (p. 25); a 14 dB notch leaves one quarter of the mean square, half of the rms, and even 20 dB reduces the rms by only about a factor of three.

The results are set against the handbook's guidelines (pp. 40–41). Guideline 1: force limiting is for highly resonant items, not for items with Q below 2; a C below 2 corresponds to a load-to-source mass ratio above about 0.5 and needs analysis or data to support it. Guideline 3: at least a 6 dB resonance should remain after notching, a notch deeper than 14 dB calls for the counsel of colleagues or a supervisor, and force limiting is not a way to lower an acceleration specification that is too high. Sweitzer's criterion (Appendix D) notches by half the resonance in decibels, which is C² = Q in the semi-empirical method.

For a random test the tool also prints the rms acceleration of the center of gravity by Miles' equation without force limiting, Eq. 5, σ = √[(π/2) Q f0 SAA(f0)], and with the semi-empirical limit, Eq. 6, σ′ = √[2 C f0 SAA(f0)] (pp. 39–40). Both are single-mode approximations. The handbook recommends a peak factor of 4.5 to 5 rather than 3 when such an rms is compared with a quasi-static limit load.

What it assumes

The acceleration specification is a correct envelope of the interface acceleration peaks in flight; force limiting does not repair an acceleration specification that is too high or too low. The flight force and acceleration peak at the same coupled resonances. One mode of the source and one of the load matter in each band. The two-degree-of-freedom limits assume the worst tuning of the load against the source. The asymptote option assumes smooth power-law residual masses and puts some modal mass in every band above the break. The notch estimate treats the fundamental mode as a single-degree-of-freedom resonance with the entered Q and mass fraction, under an input that is flat across the resonance. Total in-axis force is limited; moments and cross-axis forces are not considered.

When it misleads

A value of C² borrowed from another program is only as good as the similarity of the two configurations: the handbook's flight values near 2 come from a 65 lb instrument on a spacecraft panel and a 230 lb canister on the shuttle sidewall, its survey values (most below 5) from mounting structures deliberately more flexible than the items, and a light item on heavy structure needs a much larger value (Figure 8). Where the flight force and acceleration peak at widely different frequencies, as in the GLAST spacecraft data of handbook section 7.3 (force at launch-vehicle modes, acceleration at spacecraft modes), the semi-empirical and simple methods do not apply. Asymptotic masses are a planning estimate; the limit is to be checked against the apparent mass measured in the low-level run. The tool's estimate of the asymptotic mass for the simple model is a stand-in for a smoothed measurement. The band values of the two-degree-of-freedom limits can change abruptly where the first modal mass of the load appears; a force specification is drawn smoothly through them. The complex-model limit is sensitive to Q when the load is much lighter than the source. The notch estimate ignores fixture mass above the gages (the handbook: add the adapter mass times the acceleration specification to the limit when the adapter is heavy), gage calibration (the gages typically read 80 to 95 percent of the force because the preload bolts carry the rest) and the controller's behavior; the actual notch is whatever the measured force demands. The impedance method of handbook section 5.6 (which needs the free acceleration spectrum of the source, not a test specification), finite-element force limits and the quasi-static limit on the measured force are not computed here.

Checks

The complex-model calculation reproduces 1,509 of the 1,512 cells of the handbook's Tables 2 to 4 that have a source modal mass (Q = 50, 20 and 5; the same tables as Tables 1 to 3 of RP-1403) exactly as printed; the three others are one cell of each table (α1 = 8, α2 = 0.5, μ = 10: printed 6, 6 and 5, calculated 9.3, 9.3 and 7.5, the printed values being those of the last tuning step alone). The handbook's impedance-method example of two identical oscillators with Q = 50 (Figure 9, pp. 35–37) is reproduced to every printed digit: 51.76 and 84.08 for the interface acceleration and force at the frequency ratio 0.62, 8.52 and 5.25 at 1.62, a load apparent mass of 1.62 and the ordinate 2.64. Those printed values are the values on a frequency grid of 0.01; the peaks themselves, at 0.618, are 58.6 and 94.7, and the ordinate there is 2.62, the value of Eq. B2. Chart readings of Figures 4 and 8, the mass ratios of Appendix D and Guideline 1, and the arithmetic of the reported C² values (pp. 43–49) are checked as well.

With the residual-mass option the tool reproduces the spreadsheet example of revision B's Appendix C (CRIS instrument on the ACE spacecraft, Q = 20): all fifteen printed simple-model limits and all fifteen semi-empirical limits to the printed digits. That spreadsheet interpolates the whole-number table for the complex model (its printed limits correspond to normalized values 6.00, 6.00, 6.61, 6.81, 6.41, 6.16, 6.32 and 6.36 from 200 to 1,000 Hz), where the direct calculation gives 6.12, 6.53, 6.82, 6.78, 5.87, 5.95, 6.04 and 6.22; the tool's complex-model limits therefore differ from those printed ones by up to 9 percent, in both directions. Both models are also checked against an independent eigenvalue and frequency-response solution of the mass, stiffness and damping matrices, and the notch formula against numerical integration. Revision C prints no worked force-limit calculation with the asymptotic masses; the default simple-model limits have no printed counterpart.