Operational Modal Analysis

Operational modal analysis (OMA) estimates modal properties from structural response measured during normal service or ambient excitation. It is called output-only because the excitation forces are not measured directly. The engineer records responses such as acceleration, velocity, displacement or strain and identifies modes from the statistical structure of the response data.

For a response vector \mathbf{y}(t) measured at several locations, OMA works with response spectra, correlations or state-space models derived from output data. The desired modal properties are:

where f_n is modal frequency, \zeta_n is damping ratio and \boldsymbol{\phi}_n is mode shape for mode n.

Engineering Role

OMA is useful when controlled excitation is difficult, unsafe, too expensive or unrealistic. Examples include bridges under traffic, floors under occupancy, towers under wind, ships under wave and machinery excitation, aircraft during ground or taxi conditions, offshore structures under environmental loading and large equipment during operation.

The method is valuable because it can identify modal behaviour under real boundary conditions and operating states. It is also limited because the input is unknown. The engineer must ask whether the ambient excitation is broad enough, stationary enough and independent enough to reveal the modes of interest.

Common outputs include:

• estimated natural frequencies;
• damping ratios;
• relative mode shapes;
• stability diagrams or pole clusters;
• singular-value spectra from frequency-domain decomposition;
• trend indicators for structural health monitoring.

Output-Only Assumptions

OMA does not produce a force-normalized frequency response function. Since input force is not measured, mode shapes are usually relative rather than mass-normalized unless additional information is supplied.

Many OMA methods assume that the unknown excitation is broadband, approximately random and not strongly correlated between input locations. If a dominant harmonic source controls the response, the method may identify the forcing signature rather than a structural mode.

A practical OMA review therefore separates:

• structural modes;
• harmonic operational orders;
• environmental or traffic loading changes;
• sensor noise;
• nonstationary operating conditions;
• boundary-condition changes.

Worked Example: Check Frequency Resolution for Ambient Data

A bridge is monitored with accelerometers for:

The basic frequency resolution of one full record is approximately:

Two candidate peaks appear at:

and:

Their separation is:

The separation in frequency bins is:

Engineering comment: the record length is adequate to separate the two peaks in frequency. That does not prove they are two structural modes. The engineer still checks mode-shape consistency, damping estimates, singular-value separation, sensor layout, traffic or machinery orders, wind conditions, stationarity and repeatability across records.

If a shorter record of:

were used, the resolution would be:

The same two peaks would be only:

bins apart, making damping and close-mode separation much less reliable.

Distinction from Related Terms

Operational modal analysis is not conventional experimental modal analysis with measured force. Conventional modal testing often uses impact hammers or shakers to estimate FRFs. OMA estimates modes from response-only data.

Operational modal analysis is not a ground vibration test. A ground vibration test usually uses controlled excitation and a planned aerospace test configuration. OMA may be used in aerospace contexts, but it is defined by output-only identification under operational or ambient excitation.

Operational modal analysis is not a frequency response function. OMA can use response spectra or correlation functions, but without measured input it does not produce a force-normalized FRF.

Operational modal analysis is not structural health monitoring by itself. OMA may provide modal features used in monitoring, but health assessment also needs baselines, environmental normalization, damage hypotheses, thresholds and maintenance decisions.

Operational modal analysis is not peak picking alone. Peak picking can be one simple frequency-domain method, but robust OMA often uses stabilization diagrams, frequency-domain decomposition, stochastic subspace identification or repeated tests.

Validation and Common Mistakes

A defensible OMA study states sensor locations, measured response quantity, sampling rate, anti-alias filtering, record length, operating conditions, environmental conditions, preprocessing, identification method, mode-shape normalization, uncertainty, repeatability and how harmonic forcing was screened.

Common mistakes include:

• treating a machine order, traffic rhythm or electrical harmonic as a structural mode;
• using too few sensors to distinguish mode shapes;
• estimating damping from short or nonstationary records;
• comparing OMA mode shapes with finite-element modes without checking coordinate mapping and normalization;
• ignoring temperature, boundary-condition, load or humidity effects on modal frequency;
• assuming output-only data can recover absolute force-normalized FRFs;
• accepting a mode because it appears as a spectral peak without checking spatial consistency and repeatability.