OptimalTREND™ Wiki / Condition monitoring
Technical
Condition monitoring
Condition monitoring is how predictive maintenance sees a failure develop. At practitioner level it is about matching the technology to the failure mode and reading the signature it leaves. OptimalTREND™ is sensor-agnostic and reads these signals from the infrastructure you already run.
The technologies and what they catch
| Vibration analysis | Rotating-machinery faults: imbalance, misalignment, looseness, rolling-element bearing and gear defects, resonance and electrical. The broadest and usually the earliest mechanical indicator. |
|---|---|
| Acoustic emission / ultrasound | Very early bearing and gear distress, lubrication starvation, valve and steam-trap leaks, electrical partial discharge. Highest-frequency, earliest on the P-F curve. |
| Oil and wear-debris analysis | Gearbox, hydraulic and bearing wear: spectrometric wear metals, ferrography, viscosity, TAN and TBN, water and particle count to ISO 4406. Identifies the wearing component and the mechanism. |
| Infrared thermography | Electrical connections and loading, mechanical friction, insulation and refractory, with quantitative delta-T criteria. Generally a later-stage indicator. |
| Motor current signature analysis | Broken rotor bars, air-gap eccentricity, stator faults and load anomalies, sensed electrically from the supply without a transducer on the machine. |
Vibration analysis, in depth
A vibration signature is read four ways, and a Category-level analyst uses all four together:
- Time waveform for impacting, truncation and modulation the spectrum hides: bearing impacts, cracked gear teeth, looseness.
- FFT spectrum where frequency identifies the fault and amplitude its severity.
- Phase to separate faults that share a frequency: imbalance, misalignment, bent shaft, looseness and resonance all sit near 1x and 2x and are told apart by phase.
- Envelope, or demodulation, to lift the repetitive bearing and gear impacts out from under the low-frequency machine vibration, which is what buys the early warning.
Analysis views, and how to read them
The four reading methods above are presented as a set of displays, and a Category-level analyst moves between them because each one makes a different fault obvious. Here is what each view shows, how it is read and how OptimalTREND™ presents it.
| Time waveform (TWF) | The raw signal, amplitude against time, usually acceleration. Read it for what averaging hides: periodic impacts from a bearing or gear defect, amplitude modulation, waveform truncation from looseness or rub, and one-off transients. Crest factor, the peak divided by the RMS, rises with impacting well before the overall level does. Essential on slow shafts, where a spectrum has too few cycles to work with. OptimalTREND™ keeps a synchronised waveform beside every spectrum and trends peak and crest factor. |
|---|---|
| FFT spectrum | The waveform transformed from the time domain into amplitude against frequency. Frequency names the source, 1x imbalance, 2x misalignment, the non-synchronous bearing frequencies, gear-mesh, and amplitude sizes the severity; harmonics and sidebands refine the call. It only separates two close peaks when the lines of resolution across the Fmax are fine enough, and a Hanning window suits steady running. OptimalTREND™ auto-labels 1x and the computed bearing and gear frequencies, sets band alarms on them and trends each band. |
| Waterfall and cascade | A stack of spectra. Stacked over time it is a waterfall, showing a fault emerge and grow; stacked over speed on a run-up or coast-down it is a cascade, or Campbell diagram, which separates speed-following orders from fixed structural resonances and exposes critical speeds. You read the ridge lines across the stack. OptimalTREND™ assembles the waterfall automatically from stored history, so a rising peak is obvious at a glance, and captures the cascade where a tacho is present. |
| Cepstrum | The spectrum of the log spectrum, on an axis of quefrency in milliseconds. It collapses a whole family of evenly spaced harmonics or sidebands into a single line, which is exactly what gear sidebands, bearing harmonic families and blade-pass groups produce, so a periodicity that clutters the FFT becomes one trendable number. OptimalTREND™ applies it to gearboxes and bladed machines to detect and trend the sideband families an ordinary spectrum obscures. |
| Envelope, or demodulation | A band-pass filter is placed around a high-frequency resonance that the small repetitive impacts excite, then the envelope of that band is itself transformed, exposing the low repetition rate of the impacts. This is the earliest reliable bearing and gear-tooth detection, at stage 1 to 2, long before the defect frequencies reach the velocity spectrum. Branded variants include PeakVue, spike energy and enveloped acceleration. OptimalTREND™ runs envelope analysis on rolling-element assets and alarms on the enveloped defect frequencies. |
| Order analysis, or order tracking | The frequency axis is rescaled from Hz to shaft orders using a tacho or key-phasor, so the orders stay in fixed bins as speed varies and rotational content is cleanly separated from resonance. Indispensable on variable-speed drives, where an ordinary spectrum smears. OptimalTREND™ order-tracks tacho-equipped variable-speed machines, so their alarms stay valid across the speed range. |
| Orbit and shaft centreline | On fluid-film bearings fitted with X and Y proximity probes, the two probes plot the shaft path within its clearance. The orbit shape and its direction read out oil whirl and whip, rubs, misalignment and preload; the shaft centreline plot shows the average position and attitude angle as load or speed changes. OptimalTREND™ renders orbit and centreline where proximity probes exist, typical of turbomachinery. |
| Phase, and operating deflection shape | Phase is the timing of the vibration against a reference. It separates faults that share a frequency, imbalance, misalignment, bent shaft, looseness and resonance all sit near 1x and 2x, and animated across many points it becomes an operating deflection shape that shows how the whole machine is moving. OptimalTREND™ carries phase alongside amplitude, so a 1x peak is diagnosed rather than merely flagged. |
| Bode and Nyquist (transient) | On a run-up or coast-down, 1x amplitude and phase are plotted against speed, as a Bode plot or a polar Nyquist. A peak with a phase shift through it locates a critical speed or resonance, and the same data drives field balancing. OptimalTREND™ captures the start and stop transient on critical machines to place their resonances and confirm balance response. |
Why the settings matter. The same fault reads clearly or vanishes depending on how the data is captured. Fmax is set to reach the highest frequency of interest, the bearing high-frequency region or about 3.25 times gear-mesh. The lines of resolution must be fine enough to split running speed from a nearby bearing frequency, commonly 1600 to 6400 lines. Spectral averaging lifts the repeatable signal out of noise. A Hanning window suits steady running, a uniform window suits transients. The time waveform needs several shaft revolutions, more at low speed. OptimalTREND™ stores these settings with every reading, so a measurement is repeatable and comparable over time.
These are not separate tools to buy, they are views of one signal. OptimalTREND™ computes them from the raw or high-frequency stream, sensor-agnostic, from the sensors, historian or SCADA already in place and added sensing only where a critical asset is blind. It derives the fault frequencies from the asset kinematics, places the band alarms, trends the overalls and the bands, and shows the supporting plot next to each diagnosis, so an analyst sees the evidence behind the alarm, not just a red light.
Fault signatures
| Imbalance | Dominant 1x radial, steady with load; about 90 degrees phase between horizontal and vertical on the same bearing. |
|---|---|
| Misalignment | High 1x and 2x with significant axial vibration, often over half the radial; about 180 degrees phase across the coupling. |
| Mechanical looseness | A raised noise floor and a train of harmonics (2x, 3x and up), sometimes half-order subharmonics; often directional. |
| Rolling-element bearing | Non-synchronous defect frequencies (BPFO, BPFI, BSF, FTF) with harmonics and sidebands; earliest in the envelope and at the bearing high-frequency resonances. |
| Gears | Gear-mesh frequency (teeth times shaft speed) with running-speed sidebands; hunting-tooth and 2x GMF for specific defects. |
| Electrical (induction motor) | 2x line frequency, and pole-pass-frequency sidebands around 1x for rotor-bar and eccentricity faults. |
| Oil whirl and whip | Subsynchronous 0.38 to 0.48x whirl; whip locks onto a rotor natural frequency as speed rises. Fluid-film bearings. |
| Resonance | Amplified response near a natural frequency with an approximately 180 degrees phase shift through it; confirmed by bump test or coast-down. |
Bearing defect frequencies
BPFO = (N/2) · (1 − (Bd/Pd)·cosφ) · fr
BPFI = (N/2) · (1 + (Bd/Pd)·cosφ) · fr
BSF = (Pd / 2·Bd) · (1 − ((Bd/Pd)·cosφ)²) · fr
FTF = (1/2) · (1 − (Bd/Pd)·cosφ) · fr
N rolling elements, Bd ball diameter, Pd pitch diameter, φ contact angle, fr shaft rotational frequency. These are non-integer multiples of running speed, which is why they stand out from imbalance and misalignment, and why OptimalTREND™ places band alarms on them.
Severity and the four bearing stages
Overall vibration is assessed against ISO 20816 (broadband velocity, mm/s RMS) in zones A (new), B (unrestricted long-term operation), C (unsatisfactory, short-term only) and D (damage). A developing bearing fault runs four stages:
- Stage 1 ultrasonic and high-frequency envelope only; weeks to months of life left.
- Stage 2 defect frequencies appear at the bearing natural frequencies with envelope harmonics.
- Stage 3 BPFO, BPFI and BSF harmonics and sidebands grow in the velocity spectrum; plan the replacement.
- Stage 4 a raised broadband floor, random high frequency, a rising 1x and finally heat and noise; failure is imminent.
Sensor-agnostic and brownfield-ready
OptimalTREND™ reads these signals from existing sensors, the historian or SCADA of any brand, and adds dedicated sensing only where a critical asset is blind. The analysis above runs on the data you already have.