The technology
Two physical signals.
One closed loop.
Dr. Elevage does not read the controller, watch the car with a camera, or depend on a manufacturer’s protocol. It measures vibration and acoustic signals generated during normal operation, detects sustained deviations, points technicians to suspected components and inspection locations, and preserves a record for before-and-after verification.
01 · The loop
Sense → Learn → Detect → Verify.
Each step feeds the next, and verified repair outcomes feed back into the system. As the operating record grows, that feedback helps refine future assessments across your fleet.
02 · Signal chain
From the car top to the report, without connecting to the control system.
Nothing in this chain requires an integration with the elevator’s own control system. That single design decision is why the product is not tied to any controller brand, and why a rollout is a maintenance task rather than an IT project.
03 · Dual modality
Why sound and vibration, and not just one of them.
Vibration monitoring alone catches mechanical anomalies well. Adding acoustics exposes a class of early-stage faults that never make it into the structure as measurable movement — and, just as usefully, it lets the system tell two different faults apart when their vibration signatures look alike.
What vibration is good at
Vibration reveals many conditions that affect the structure: rotating imbalance, bearing-race defects, guide-rail joints and misalignment, changing loads in hoist ropes and compensation components, and ride quality. Three linear axes show direction, while three rotational axes add information about twist and sway that a single-axis sensor cannot capture.
What acoustics adds
Friction, air and contact events that barely register as movement: brake pad scrape, door-operator gear whine, compensation chain or rope rubbing the shaft wall, early lubrication loss. These show up as a change in spectral content long before they show up as amplitude — which is precisely why an experienced technician listens before they look.
One modality tells you something changed. Two modalities, cross-referenced against the same trip, tell you what changed and where — and that is the difference between an alarm and a work order.
04 · Per-unit baselines
An old elevator is not a broken elevator.
This is the single most common reason condition monitoring fails on real building stock: a fixed threshold flags every aging unit and nothing else, so the alerts get ignored within a month. Dr. Elevage sets the threshold from the unit’s own history.
Learn the resting state
As operating data accumulates, the system builds a characteristic baseline for each elevator across load, direction, speed and floor — not one number, but a family of them. While the baseline is developing, the system can reference comparable units and duty cycles; once it stabilizes, changes are evaluated primarily against that elevator’s own operating history.
Compare like with like
A full car going up at 1.0 m/s is only ever compared with other full cars going up at 1.0 m/s. Load and direction are controlled for, so real drift is not buried in normal variation.
Flag the trend, not the spike
One noisy trip is a passenger jumping. A sustained directional drift across hundreds of trips is a component. The system waits for a persistent pattern.
05 · Fault signature library
What the system is actually looking for.
A representative extract from the library. Each entry pairs a physical failure mode with the signature it produces, the modality that tends to reveal it first, and the way the change typically appears in the data.
| Failure mode | Signature | Seen first by | Detection pattern | Consequence if missed |
|---|---|---|---|---|
| Traction machine bearing wear | Rising high-frequency amplitude, then broadening sidebands around the running frequency | Vibration | Progressive amplitude trend | Unplanned shutdown |
| Guide-rail misalignment / joint step | Repeating lateral impulse locked to a fixed floor position, identical every trip | Vibration | Position-linked pattern | Ride complaints, accelerated shoe wear |
| Door operator degradation | Longer open/close cycle, gear whine climbing in the 2–6 kHz band | Acoustics + door zone | Cycle-time and band trend | Entrapment — doors are the single largest cause |
| Compensation chain or rope / traveling cable | Broadband rub audible only in a specific travel band, absent elsewhere | Acoustics | Travel-band pattern | Cable damage, noise complaints |
| Brake pad wear / drag | Scrape signature at start and stop, plus a shift in the deceleration curve | Acoustics + speed | Start/stop signature trend | Loss of braking margin |
| Counterweight guide-rail fault | Intermittent vibration with no car-side source | Vibration | Event-linked pattern | Hard to find by hand; often misdiagnosed |
| Roller guide / guide shoe wear | Gradual rise in lateral RMS across the whole travel, no fixed position | Vibration | Gradual whole-travel trend | Ride quality decline |
| Lubrication loss | Spectral content shifts upward before any amplitude change is measurable | Acoustics | Spectral shift before amplitude | Accelerated bearing wear, avoidable at this stage |
Representative extract. The full library covers 50+ fault signatures, built on traction elevators, and is extended with every verified repair outcome.
06 · Technical FAQ
The questions engineers ask.
How do you separate the elevator’s signature from passenger noise?
Two ways. First, the analysis is trip-scoped and conditioned on load, direction and speed, so a one-off event does not move the baseline. Second, faults have structure — a bearing defect repeats at a frequency tied to shaft rotation, a rail joint repeats at a fixed floor position. Passengers do not repeat. The alert requires a sustained, structured deviation across hundreds of trips, not a loud afternoon.
Does it need to connect to the elevator controller?
No — and this is deliberate, not a limitation. The sensor measures physical vibration and sound from outside the system. No controller bus, no manufacturer protocol, no OEM permission, no firmware on the elevator. That is why it can be deployed across a wide range of traction-elevator brands and ages, with special elevator types, site conditions and duty profiles assessed individually before deployment.
If you never touch the controller, where do speed, load and floor position come from?
Speed, load and floor position are derived by our own algorithms from the sensor data itself — the system never connects to the control system, and takes no signal from it.
What happens when the site loses network coverage?
Data is buffered locally and uploaded when connectivity returns; the record is written by the device at the point of capture, not assembled at the point of upload, so the trend stays intact. The Portable Inspector supports fully offline collection by design — basements, machine rooms and shafts with no cellular coverage are normal working conditions, not exceptions.
How long before the system produces useful output on a new unit?
Position-related anomalies — such as rail-joint impacts, misalignment and door-zone anomalies — can appear in the first passes because they are tied to a fixed location in the shaft rather than a long-term trend. Trend-based warnings require the unit’s baseline to stabilize across its normal duty cycle first; in our traction-machine case, the developing anomaly became clear over a two-month trend.
Can the data be edited after the fact?
Not by the party being assessed. Operating records are written by the device and uploaded as recorded; the console gives no one an interface to edit them, and there is no manual check-in, uploaded photo or technician-entered field anywhere in that chain.
What about hydraulic, screw-driven elevators or escalators?
Our focus today is traction elevators, and that is where the fault library is deepest. Hydraulic, screw-driven and other types of elevator need targeted adaptation. If you have escalators in mind, we would also be glad to hear from you and explore the application together.