Nothing on this page
was drawn by hand.
Three demonstrators, all running before you touched anything. Every waveform is synthesized in your browser from a physical model, every spectrum is a real FFT of those samples, and every frequency label is computed from bearing geometry — the same formulae that generate TCAT's assessment cases. Hover any plot for a cursor readout.
It is already running
Each demonstrator starts on its own and cycles through its states, so the difference between a healthy machine and a failing one plays out in front of you. Take manual control any time — the cycling stops and waits.
The signals are synthetic
Deliberately. A synthesized case has an exact known answer, which is the only way to mark a diagnosis objectively. Nothing here is a recording of a real machine, and nothing here is presented as one.
The maths is not decorative
Band-pass, Hilbert transform, envelope, FFT, single-bin phase — run in the page, on the samples you are looking at. Change a state and the numbers move because the transform re-ran, not because a caption changed.
Watch a bearing fail — and watch the raw spectrum miss it.
An outer-race defect on a 6205-type bearing at 1500 rpm, in three states. The overall level barely moves between healthy and early. The raw spectrum has no line at BPFO. The envelope spectrum has one, with harmonics. That gap is why enveloping exists, and it is the single most useful thing a Category II analyst knows.
Bearing fault · outer race · envelope analysis
The record itself. Green ticks mark the nominal BPFO spacing — the impacts jitter about them, which is exactly why a bearing line smears while a shaft order stays sharp.
Raw spectrum from 200 Hz up, so the shaft orders do not set the scale. The impact energy is not missing — it is parked up at the housing ring-down, thousands of hertz above the defect rate that caused it.
The same raw spectrum, zoomed to where BPFO, BPFI, BSF and FTF would appear. In the early state there is nothing there to find.
Band-pass the ring-down, take the Hilbert envelope, transform again. BPFO and its harmonics, from the same samples.
The machine the record came from. The flash at the housing is an impact as the rolling elements pass the defect.
Two details worth naming, because they are what separates a competent analysis from a lucky one. First, the demodulation band is found, not chosen: the page scans candidate bands and keeps the one whose band-passed envelope has the highest kurtosis — the most impulsive one, which is not always the one with the most energy. Second, every defect frequency on the plots is computed from the bearing's geometry at the running speed, not typed in. Change the speed and they all move, in the ratios the kinematics fix.
Fig. 1 — where BPFO, BPFI, BSF and FTF come from. Four contact geometries, four rates, one set of formulae.
Fig. 2 — the envelope chain the panel above actually executes: band-pass, demodulate, low-pass, transform.
An operating deflection shape is a machine caught in the act.
Amplitude alone tells you something is wrong. An ODS tells you what is moving against what. Each station on this train moves at its own measured amplitude and its own measured phase, so the animation is not an impression of shaking — it is the deflection shape, and it changes character completely between the three states.
Operating deflection shape · motor–pump train
Reference: a beam's second bending mode. When an ODS looks like a mode shape — antinodes, a dead node, a 180° flip across it — the rotor is not the problem. The structure is.
The phase column is measured, not asserted: each station's signal is synthesized, then its amplitude and phase are read back out of the samples by a single-bin transform. The table's last column is the one that decides — a station 180° from the reference is moving the other way, and the row turns. In the soft-foot state exactly one foot and the plate under it do that. In the resonance state half the machine does.
Phase is the reading that settles the argument.
Unbalance, misalignment and looseness all raise 1×. The spectrum narrows it; the phase relationships close it. The needles below all turn at running speed, each trailing by its own measured lag against the tacho reference — so the 90°, the 180° and the refusal to repeat are things you watch, not things you are told. The survey is re-taken every couple of seconds, live.
Phase survey · unbalance vs misalignment vs looseness
The spectrum at 2V. It narrows the field — it does not close it. Two of these three states can look like this.
Shaft centreline orbit for the same state: a circle for a rotating force, a figure-of-eight when it reacts twice a turn, truncated when something is loose.
The last chip is the one people underuse. A phase reading that will not repeat is not a bad reading — it is the finding. Looseness does not re-seat identically every revolution, so the number moves between acquisitions, and that instability is diagnostic in a way no single number on a spectrum is. Watch the row of successive 2V readings above: steady for unbalance, wandering for looseness.
Synthetic on purpose, and computed on purpose.
Every signal on this page was generated to specification by TIERA's own case generator — fault class, severity, speed and seed — and none of them is a recording of a real machine. That is the point, not a compromise: a synthesized case has an exact known answer, so a diagnosis can be marked against ground truth rather than against an opinion, and the same case can be reproduced identically for every candidate. The models are the published kinematics and the ordinary mechanics of impacts, resonances and non-linear joints; the same generator produces the cases used in TCAT assessment.
Nothing here is fetched, and nothing here is pre-rendered. The page ships a signal generator, an FFT, a band-pass, a Hilbert transform and the bearing kinematics, and your browser runs them. If you scrub a state, the numbers change because the transform ran again.
What this page does not do is stand in for the assessment. It is a demonstration of the physics the syllabus is built on. The credential is decided by graded diagnostic tasks — see how assessment works.
TCAT is a TIERA-issued credential aligned with the ISO 18436-2 body of knowledge. It is not an accredited ISO certification, TIERA is not an ISO certification body, and TCAT training hours do not count towards the formal training an accredited ISO 18436 certification requires.
If what you just watched felt unfamiliar rather than obvious, you are not TCAT's audience yet — and that is a fine place to be. TCAT assesses people who already work with vibration data; it does not teach the subject from zero. The honest starting point is the free TIERA 101 primers: self-paced, beginner-first, no cost, and they issue a completion badge when you finish. Go there first if that is where you actually are — nothing about TCAT is a reason to skip them, and nothing you do there is required before you come back.
This is the level the assessment is written at.
If reading these three panels felt like your job, TCAT is aimed at you. If it felt like a foreign language, start with the free primers first.