Every machine leaves your factory with proof.
Bind a test run to a work order or serial number, capture the condition parameters that matter while it runs, and issue a test certificate the customer can trust — and that you can produce again three years later when a warranty claim arrives.
The test report is the last thing you build, and the least engineered.
Most OEMs run genuinely rigorous tests and then record them on a printed sheet with a signature. The testing is world class. The evidence is a photocopy in a file cabinet.
"It was fine when it left here"
A gearbox fails at eleven months and the customer says it was never right. Without recorded test data you are arguing from memory against a purchase order. With it, you attach the run and the conversation ends.
Readings taken by hand, once
A technician notes vibration at three points at one moment. Everything between those points — the ramp, the transient, the drift at temperature — is unrecorded and unrecoverable.
Serial number 4471, three years on
Finding one machine's test record means someone walking to a cupboard. If the file is missing, the record does not exist — and neither does your defence.
Scan, run, certify.
The operator's job stays almost unchanged. Everything else happens because the bench is instrumented.
Bind the run
Scan the work order or type the serial number at the bench. The run is now attached to that specific unit, its model, its customer and its test profile — before a single reading is taken.
Run the profile
Ramps, hold points, load steps and durations defined per model. The system tracks adherence: whether the unit actually held 1,480 rev/min for the full twenty minutes, not whether someone remembered to.
Capture continuously
Vibration with live FFT, temperatures, speed, torque, pressure, flow, current and power — every second of the run, not three spot readings.
Issue the certificate
Pass or fail evaluated against your limits and ISO zones, then a test certificate generated against the serial number — stored on your server, retrievable in seconds, forever.
What the customer receives.
A certificate that says more than "tested OK" — because the data behind it exists, it can carry actual evidence rather than an assertion.
Search by serial, not by cupboard
Every run stays queryable by serial number, work order, customer, model, date or result. A claim on a three-year-old unit is a search box, not an afternoon.
The full run behind every line
The certificate summarises; the underlying run is retained in full. When a customer's engineer asks what the spectrum looked like at 100% load, you show them.
On your server, not a vendor cloud
Test results describe your product's real performance — commercially sensitive by definition. MachineWise is on-premises, so that data never leaves your factory.
If it can be run on a bench, it can be certified.
The platform is equipment-agnostic — you define which parameters matter and what the limits are. These are the product families we see most.
Tri-axial vibration with live FFT and envelope analysis · bearing and oil temperatures · speed and torque · pressure and flow · three-phase current, voltage and power · noise level where instrumented · ambient conditions · run duration and profile adherence. Limits are yours to define — per model, per customer specification, or against ISO 10816 / 20816 zones.
Four returns, only one of which is the certificate.
Claims you can answer
A recorded run at dispatch shifts the burden of proof. Most disputed claims resolve the moment the data appears, and the ones that do not are at least argued on evidence.
Benches that wait, measured
Test bed utilisation is usually the quiet constraint on dispatch. Run counts, queue depth and idle time per bench turn that into a number you can plan against.
Patterns across the population
When every unit of a model is measured the same way, drift becomes visible — a supplier's bearing batch, a fixture wearing, an assembly step changing. That is design feedback no field report gives you.
A certificate competitors cannot match
Sending a real test record with every unit is a visible quality signal in a market where most competitors send a stamped sheet. It justifies price, and it wins audits.
The same platform runs the rest of your plant.
The sensor inventory, by what it tells you.
| Parameter | Typical sensing | What it establishes on a test bed |
|---|---|---|
| RPM / speed | Encoder, proximity or tacho | The independent variable for almost every other reading. Defect frequencies, power and flow all scale with it. |
| Torque | In-line torque transducer or dynamometer | With speed, gives power. The basis of any performance curve. |
| Vibration | Tri-axial accelerometer | Bearing and gear condition, imbalance, misalignment, and impact events during the run. |
| Acceleration / shock | High-rate accelerometer channel | Transient events a slower channel would average away — impacts, engagements, sudden load steps. |
| Temperature | Thermocouple, RTD, or IR where contact is impractical | Bearing, oil, winding, coolant and exhaust temperatures, and the rate at which each stabilises. |
| Pressure | Pressure transducer | Hydraulic, lubrication, air and process pressures — and their stability under load. |
| Flow | Flow meter | Coolant, lubricant or fuel flow. With time, gives consumption per test. |
| Current / voltage / power | A smart energy meter on the rig supply | Electrical load through the run, plus efficiency where mechanical output is also measured. |
| Level | Level transmitter or float | Reservoir and tank levels — a slow leak is visible across a long endurance run and nowhere else. |
| Humidity | Humidity transmitter | Ambient conditions, for correcting results and for demonstrating the test environment was within specification. |
| Moisture | Moisture sensor in oil or air line | Contamination, which changes both the result and the article's condition afterwards. |
| pH / conductivity | pH or conductivity probe | Coolant, dielectric and process fluid condition on rigs where fluid chemistry affects the outcome. |
| Noise | Sound level meter | Acoustic performance where it is part of the acceptance criteria. |
| Displacement / position | LVDT, linear encoder | Deflection, backlash, wear and end-float under load. |
| Strain | Strain gauge bridge | Structural loading on the article or the fixture, where that is what the test is about. |
| Digital states | Guard, interlock and mode signals | Guard, interlock, mode, start and abort events — the context that makes an analogue trace interpretable. |
Every channel is timestamped on the same clock, which is the point. A temperature rise that coincides with a torque step is a finding; the same two readings on separate systems with separate clocks are two files nobody reconciles.
Read it as a chain rather than a list: each stage depends on the one before. A report is only as trustworthy as the traceability behind it, and traceability is only meaningful if execution captured the identity of what was being tested at the moment it was tested.
Channels are configured per article type, not per bench, because the same rig tests different products with different acceptance criteria. Sampling rate is set per channel: a vibration channel looking for impacts and a tank-level channel watching for a slow leak have nothing in common in this respect, and a single platform-wide rate is wrong for both.
Where the specification defines a profile — ramp rates, hold points, load steps, durations — it is entered once and reused, so every article of that type experiences the same test and any deviation is visible rather than inferred.
| Captured | Why it matters |
|---|---|
| Test ID | A unique handle for this run. Everything else hangs off it. |
| Product and serial number | Bound at the start by scan or selection. Without it you have a test result attached to nothing. |
| Test recipe | Which specification was applied, and its revision — the question an auditor asks first. |
| Test sequence | The steps as run, in order, with their actual durations against the specified ones. |
| Timestamps | Every step and every reading on one clock, so cross-channel correlation is possible afterwards. |
| Operator | Who ran it and who witnessed it, where witnessing is part of your process. |
Pass or fail against your criteria
Per channel, per step — a limit that applies during a load hold need not apply during the ramp into it. Criteria are yours; we do not impose thresholds.
Within the run and across the population
A reading drifting through a long run is a finding. The same article compared against every previous unit of that model is a different and often more useful one.
Behaviour unlike the model's norm
Once a population exists, an outlier is visible without anybody having set a threshold for it — which catches the failures nobody wrote a limit for.
Did the test actually run as specified
Whether hold points were held and ramps were ramped. 'Ran the 100% step for thirty minutes' becomes a verified fact rather than a note.
Traceability binds product, test, measured parameters, operator and result into one record, retrievable by serial number, work order, customer, model, date or outcome. Three years later, when a warranty claim arrives, that is the difference between attaching the run and arguing from memory.
Output is a digital test report generated against the serial number, a live dashboard while the bench is running, a searchable historical archive, and alerts when a limit is breached or a long run terminates unexpectedly. On unattended endurance testing the alert matters most: a run that stops at 02:40 and is discovered at 07:00 has cost the night.
All of it sits on your own server. Test results describe your product's real performance and are commercially sensitive by definition, which is why MachineWise deploys on-premises.
Digitise one test bed and certify a real unit.
Send us one test sheet you fill in today. We will digitise that bench and issue a certificate against a real serial number, free for one test bed.
Straight answers.
What is test bed digitization?
How is pass or fail decided?
Which products can be tested this way?
Where is the test data stored?
Digitize one test bed, free, for thirty days.
Pick your busiest bench. We instrument it, run your real test profiles, and hand you certificates against real serial numbers — before you decide anything.
Book my free pilot → WhatsApp us