Workflows

A project has a type, and the type sets what the project expects: the data you bring, the objects it holds, the figures its report is built from, and how the result is judged. The seven types are described below, followed by correlation and finite element modeling, which work in any project.

Modal analysis

Import frequency response functions from a hammer or shaker survey, or compute them from the time data, fit modes to them on the fitting screen, and animate the shapes on the test geometry. The report includes the coherence and the drive-point FRFs, so a poor channel can be found before it affects a fit. On the fitting screen, a peak's Q can also be read straight from its half-power width, with a note when the peak has too few lines to read it.

Random vibration test analysis

Control PSDs averaged over the run's own frames, overlap and window, compared against the specification with its warning and abort lines, channel by channel. Octave-band levels, multiple coherence and RMS error bars in decibels are part of the same reading. Each PSD shows the degrees of freedom its averages are worth and the confidence band they give, and each channel can be read in the time domain first, its moments, tails and kurtosis checked against a Gaussian record's. A tone in the random is read in g peak off the PSD's lines, not off the height of its spike.

Shock test analysis

Detect the events in a record, filter and integrate it into velocity and displacement, compute a shock response spectrum for each channel and each event, and compare them against the required spectrum and its tolerance band. Signed error bars in decibels show which events under-tested or over-tested, and the points rule counts how many points of each spectrum fall inside the band. Before the spectrum, each record's velocity is checked for a zero shift. A classical pulse is judged against its envelope and its velocity change, and a pulse can be checked against a shaker's stroke, velocity and force ratings before it is run.

Transient environment analysis

A measured run compared against the target waveform the controller was asked to reproduce, one control channel at a time. Each playing is overlaid on the target where it actually landed, the waveform error is shown for every channel and playing, and the PSDs of both compare the levels. A run can also be judged segment by segment, each stretch by the rule for its kind, and a reference record can be checked against a shaker's ratings before its first pass.

Sine sweep test analysis

One or more tones swept across a frequency range at a controlled amplitude, so the test article sees each frequency in turn. The level each tone actually reached is extracted from the recording and compared against the level it was required to reach, tone by tone and channel by channel, with the difference in decibels. The sweep rate is checked against the modes it passes, a dwell is read against time with its build-up and ring-down, and two signature sweeps, before and after a test, are compared mode by mode.

Random and sine

A random test with one or more sine tones swept under it on the same shakers. One project holds both halves, each judged by its own workflow, and it generates two reports: a random report and a sine report.

System identification

The plant a later test will be controlled through: the frequency response of every control channel to every drive, measured from an ambient recording and a driven one. The report shows the densities of both recordings, every FRF, and the two measurements that say whether to trust the result: the multiple coherence map and the signal-to-noise against the ambient floor.

The signal-to-noise follows the standard definition, signal power over noise power, as IEEE Std 1241 states it and as Bendat and Piersol use it in Random Data. The driven recording holds the signal plus the same noise the ambient one recorded, so the signal's power is the driven power less the ambient power, and the ratio is that divided by the ambient power. The report draws it line by line for every channel, and once per channel over the whole band as bars against a 10 dB threshold, the margin at which the coherence that the same noise allows reaches 0.91. How it is computed.

Finite element model correlation

Compare test modes against a model's by the modal assurance criterion. The model's shapes are sampled at the test's degrees of freedom, the cross-MAC shows which modes pair up, matched pairs are kept with their frequency error, and the two shapes of a pair can be animated together on the structure.

Finite element models

Build a model in the app and solve its modes. A thin-walled part is built from planes of plate elements, with each bolt a tie over the elements its washer covers. A thicker part is built from solids (blocks, cylinders, sections such as I-beams, channels, tees, angles and rectangular and round tubes, and holes that go through, to a depth or with a counterbore), combined by subtracting and joining, copied with Pattern, and meshed into bricks with Mesh Solids, which meshes rings of bricks about each hole. Parts meshed apart are bolted by aligning their meshes and merging the nodes each bolt clamps. A model can also start from an imported mesh.

Supported data formats

Every format reads into the same objects, so what happens next does not depend on where the data came from. Most of them are written as well as read, and the table marks the ones that are read-only.

A universal file comes in two encodings, dataset 58 as text and dataset 58b as binary floats, and both are read and written. The binary one is about a third the size and keeps the values exactly, where text keeps five decimal places.

A Rattlesnake .nc4 is a controller's record of a run, and a project is not exported as one, since that file would describe a run that never happened. A run can be cut to a stretch of time and saved as a new .nc4, though: Save Truncated Copy… in the import window for a long run, or File → Truncate Rattlesnake File… for any run, copies every setting and channel as it is and cuts only the time data. STEP and IGES files belong to the CAD tool and are read-only on purpose. Punch and Femap writers have not been written yet; every other format that is read can also be written.

• See the figures
• Download
• Source on GitHub