- Home
- Applications
- Industrial Testing
Application · Industrial testing & validation
Motion platform for industrial testing, specified from the load case backwards.
In a simulator the motion has to convince a person. In a test it has to convince a measurement — so the number that decides the platform is not how well it moves, but how identically it moves on the tenth run.
Arrived from a vibration or shock requirement? A position-controlled platform and a shaker produce different physics — the crosswalk further down says which machine each clause belongs to, including the cases that are not us.
Motion platform for industrial testing
What sits on the platform
Eight test jobs an industrial motion platform is bought for
Each of these buys the same thing from the mechanism — a motion input the test owner can command, repeat and describe. What differs is what is being proved above the mounting plane, and that is what decides the architecture and the load class.
Sensors
Inertial sensor and IMU characterisation
The platform applies known rates and attitudes so the device is measured against a motion you commanded rather than one you happened to record.
Lets you separate bias, scale factor and cross-axis behaviour, because the input repeats closely enough to attribute the difference to the device.
Architecture 6DOF, or 3DOF where yaw is not exercised
Load class Usually the compact end, 150 – 300 kg
Control laws
Stabilisation and compensation verification
The platform supplies the disturbance a gimbal, stabilised mount or compensation head is built to reject, on command and on demand.
Lets you compare one control law against its previous version on identical input, which a sea trial or a road test cannot offer.
Architecture 6DOF, or Stewart geometry where the pivot matters
Load class 150 – 600 kg for heads and mounts
HIL
Hardware-in-the-loop motion rigs
Your real-time host commands platform motion while the article’s outputs return to the model inside the same loop.
Lets you exercise embedded software against physical motion before there is a vehicle, a vessel or an airframe to put it in.
Architecture 3DOF or 6DOF; the interface decides more than the axes do
Load class Set by the article, not by the loop
Durability
Durability under a reproduced field profile
The platform replays a conditioned field recording for as many cycles as the programme calls for, at the duty the article will actually see.
Lets you trace a failure to a load case that exists as a file, rather than to a bench nobody else can reproduce.
Architecture 3DOF where the record allows it, 6DOF where it does not
Load class 300 – 2,000 kg is the common band
Fluids & cargo
Sloshing, cargo and free-surface behaviour
The platform moves a tank, rack or container through a recorded attitude history while the contents do what they do.
Lets you judge a baffle design or a securing method against the motion that caused the problem in service.
Architecture 3DOF for roll and pitch records, 6DOF for deck motion
Load class 600 kg upward; contents count as moving load
Perception
Lidar, radar and camera under known motion
The platform carries a perception assembly through a commanded trajectory in front of a fixed, surveyed scene.
Lets you separate ego-motion from scene motion, because one of the two is a file you wrote.
Architecture 6DOF; the reference point is stated at the sensor, not the table
Load class 150 – 600 kg for sensor stacks and masts
Robotics
Robots and teleoperation on a moving base
The platform becomes the carrier motion a manipulator, docking head or teleoperated end effector has to work through.
Lets you test whether an arm keeps its tolerance when its base is not the fixed world the controller assumes.
Architecture 6DOF; combined-axis travel is the figure that binds
Load class 300 – 1,200 kg including the manipulator
Links
Antenna and communication link under carrier motion
The platform reproduces the motion of the vehicle or vessel the terminal is mounted on, while the link stays up or does not.
Lets you measure tracking loss and re-acquisition against a repeatable attitude history instead of against weather.
Architecture 6DOF, with yaw range checked against the geometry limit
Load class 300 – 1,200 kg for terminals and radomes
Your test is not one of these eight? Describe what has to be proved and we will say whether motion is the method →
What actually gets delivered
The 4-Configuration Test Bench Stack
A platform on its own is not a test bench. What leaves here is a platform plus the interfaces, the conditioned profile and the acceptance package that make it usable as instrumentation. These four cover most industrial test work; the fourth is where the geometry stops being standard.
C1
Sensor and component test cell
Typically 150 – 300 kg gross moving load
A bench-scale six-axis cell for devices that are measured while they move: inertial units, perception assemblies, small controllers and their brackets.
Inside the supply
- Platform, controller and electrical cabinet
- Interface plate machined to your bolt pattern
- Pose and profile command interface, with the convention frozen in writing
- Loaded acceptance with run-to-run spread reported
Not included: your device fixture above the interface plate, the data acquisition, and any calibration mathematics applied to the device.
C2
Hardware-in-the-loop motion cell
Load class set by the article, not by the loop
A platform configured to sit inside somebody else’s real-time loop: commands arrive from your host, state goes back, and the timing behaviour is defined rather than discovered.
Inside the supply
- Platform, controller and drives
- Command set, state packet and coordinate convention, frozen at the interface review
- Defined behaviour for late packets, aborts, resets and safe state
- Workspace and path checking before motion, with refusals reported to the host
Not included: the real-time host itself, your simulation model, and the instrumentation on the device under test.
C3
Duty and durability rig
Commonly 300 – 2,000 kg gross moving load
A platform sized against a cycle count rather than a peak. Everything that decides whether a machine survives ten thousand repetitions is chosen here, and peak force is rarely the binding one.
Inside the supply
- Actuators and thermal margin sized against your duty cycle
- Conditioned profile verified over a continuous block, not a single pass
- Service interval and wear-part list issued with the machine
- Acceptance run long enough to show the figures hold once it is warm
Not included: the fatigue criteria, the failure analysis and the test report — those need knowledge of the article, not of the machine moving it.
C4
Large-article test frame
Above 2,000 kg, geometry laid out per project
Where no reference frame fits: the article, the hall or an existing structure fixes the envelope, and the mechanism is drawn around it rather than selected from a range.
Inside the supply
- Geometry laid out around your moving assembly and load path
- Structural analysis against your load case
- Installation, commissioning and on-site acceptance
- The same identification and acceptance records as a reference frame
Not included: foundations, site power, permits and system-level safety approval for the installed cell.
Most enquiries land on C1 or C3. The two that cost projects when they are misread are C2, where the interface is the deliverable rather than the platform, and C4, where the schedule is set by geometry and analysis before anything is fabricated.
Not sure which of the four your programme is? Send the article mass and the profile and we will name it →
Which mechanism carries the bench
The platforms these benches are built on
Axis count is not a preference, and extra axes are not free — every one you add is mass on the moving assembly, and mass costs the acceleration available to the axes you actually needed. The source recording usually settles it before a price list is open.
Three axes
3DOF motion platform
Heave, roll and pitch on a serial mechanism. The right answer when the record carries its content in those three and the other three are small or irrelevant to the result.
Six axes
6DOF motion platform
All six commanded together on a parallel mechanism. Needed when the record is coupled — road load, deck motion, airframe attitude — because reproducing three of six changes the load path, not just the realism.
Hexapod geometry
Stewart platform
The same six legs, specified as a positioner. Chosen when the test is about holding and stepping between poses rather than reproducing a trajectory, or when the pivot point sits well away from the table.
The Axis-Count Read from Source Data Look at the record, not at the specification template
| What your source data contains | What to specify | Why |
|---|---|---|
| Heave, roll and pitch carry the content; surge, sway and yaw are small or irrelevant to the result | 3DOF | Three axes you can command, measure and defend beat six you cannot |
| All six axes present and coupled — road load, deck motion, airframe attitude | 6DOF | Reproducing three of six changes the load path, not just the realism |
| One axis dominant, at a frequency above the working range | Neither | This is the shaker row of the crosswalk, whatever the axis count says |
| Rotation referenced to a point well away from the platform surface | 6DOF | The reference point has to be stated, because it changes every travel figure |
The trade-off itself is worked through on 3DOF vs 6DOF, and electric vs hydraulic covers why every bench on this page is servo electric.
Have the record but not the axis decision? Send the file and we will read the axis content out of it →
Recorded during acceptance, not during a demonstration
A profile run, and the run after it
Each clip has a load on the platform and a trace on screen. A platform filmed moving with an empty table and no measurement is a marketing exercise, and a test engineer can tell within two seconds.
What was asked for, next to what happened. A conditioned customer profile running with both traces on screen, so tracking is something you watch rather than a number you take on trust.
The repeatability claim, filmed. Several consecutive runs of the same profile at the test load, overlaid, with the spread shown as it accumulates.
A segment of the real duty. An article on its fixture through part of a continuous programme, filmed long enough to show that nothing changes when the machine warms up.
Want to see a profile close to yours before you commit? Tell us the axis content and duration and we will send the nearest run →
Why a test platform is specified differently
A test needs repeatability first and absolute accuracy second
These two get bought as one requirement and they are not one requirement. A constant offset between what you commanded and what the platform did is a bias — measure it once at commissioning, write it down, and every result carries the same known shift. Scatter between one run and the next is different: it lands in every data point and no amount of post-processing takes it out.
So for a test article, run-to-run scatter is the expensive number and absolute position is the cheap one. A platform that sits consistently off nominal but returns to within a few hundredths of its own previous run gives you a usable test; one that averages out to nominal while wandering between runs does not, however good its datasheet looks.
That changes what acceptance is. We run your conditioned profile repeatedly at the test load and report the spread across those runs, rather than demonstrating a cycle we chose. It costs us more test-floor time and it is the only version of the figure you can put in a report.
Need the repeatability figure on your own profile before ordering? Send the file and we will quote the run count with it →
Stewart parallel robot finite element modal analysis showing second to fifth mode shapes and frequency amplitude distribution
The 3-Number Test Input Statement What goes on a test report, and what it means for your result
| Quantity | How it is stated | What it does to your result |
|---|---|---|
| Run-to-run repeatability | Peak deviation across consecutive runs of your profile, at the test load, measured rather than computed from encoders | Appears as scatter in every data point and cannot be removed afterwards |
| Tracking error against the commanded profile | Time-domain error band over the named profile, with the frequency range it holds across | Tells you how much of what the article experienced was actually what you asked for |
| Bias at the reference point | Measured at commissioning and recorded against the platform serial number | A known constant you can subtract once, provided it was measured and written down |
All three belong to a stated load and a stated profile. Quoted without those, they describe a machine running empty, which is not the machine your article sits on.
What arrives, and what it has to become
Recorded data does not become a test profile on its own
Road load files, deck recordings, flight logs and model output all arrive as time series, and none of them is directly commandable. Sample rates do not match the controller, integration drift walks the platform out of its workspace, and content above the platform’s range has to go somewhere. Doing that conversion badly is the most common way a good platform produces a bad test.
The 5-Step Profile Conditioning Path Every transformation is listed, none is applied silently
ST 01
Ingest
A time series with its sample rate, units and axis convention stated. Plain delimited text is always safe; other containers are confirmed per project. Missing units are the most common reason a file stalls on arrival.
Blocker: unlabelled axes
ST 02
Condition
Resample to the controller rate, filter content above the platform’s working range, and remove the integration drift that turns a velocity record into a slowly climbing position.
Filter corner is reported
ST 03
Scale
Fit the trajectory into the reachable workspace at your CG height. Where scaling is needed it is issued as a factor per axis, so you know what the article no longer sees.
Scaling is never silent
ST 04
Verify
Run the conditioned profile against the actuator envelope — stroke, velocity, force and thermal duty — before the frame is built rather than during your test slot.
Fails here, not on site
ST 05
Issue
The conditioned file is returned with every transformation listed, so your test report can state what was commanded and how it differs from what was recorded in the field.
Traceable back to source
Have a recording but no idea whether it fits a workspace? Send a sample file and we will return the conditioning report →
The part most enquiries under-describe
The fixture is part of the test, and usually the part that spoils it
We reproduce motion at the mounting plane. What your article experiences is that motion filtered by whatever stands between the plane and the sensor — and a fixture is never neutral. A compliant one is a low-pass filter that quietly removes the content you were trying to apply. A resonant one is an amplifier that adds content nobody asked for.
Both failures look identical from the control side. The platform tracks its commanded profile, the acceptance figures are met, and the article still sees something else. A test can be executed perfectly and be wrong, and the fixture is where that usually happens.
So the enquiry asks for the fixture’s mass, its centre of gravity above the interface and, where anyone has calculated it, its first natural frequency. If that frequency sits inside the band you are trying to apply, it needs to move before the test is worth running.
Fixture already built and the results look wrong? Send its mass and first mode and we will check it against your band →
Industrial motion platform with electric actuators undergoing factory acceptance testing
The 8-Line Test Definition What we need before a quotation means anything
| Line | What it changes on the platform | If it is left blank |
|---|---|---|
| Article mass and CG height above the interface | Frame section, actuator force and the achievable rotation | We size to a centred, low article and the real one costs travel |
| Fixture mass, CG and first natural frequency | Whether the band you want to apply survives the fixture | The test runs and the article sees a filtered version of it |
| Profile source, sample rate and units | The conditioning path and what can be reproduced from it | The file cannot be used until we come back and ask |
| Which axes, and whether they are commanded together | Architecture, and whether three axes are enough | Six are quoted where three would have been the better test |
| Frequency content that matters to the result | Whether this is our machine at all — see the crosswalk | The wrong class of machine gets quoted |
| Duty cycle and total cycle count | Thermal margin, actuator rating and service interval | Sized for intermittent duty, run continuously, and it runs hot |
| Host interface and data timing requirement | Command level, coordinate convention and loop budget | Integration becomes a commissioning problem instead of a design one |
| Acceptance method and tolerance | What the loaded FAT measures and who signs it | Acceptance defaults to our protocol rather than your test |
Eight lines, and most enquiries arrive with four of them. Write “unknown” on the other four rather than an estimate — an unknown gets asked about, an estimate gets built to.
Which machine your clause is describing
Which test input belongs to which machine
A motion platform is a position-controlled machine. It puts a large mass where you ask it to be, along a trajectory, at frequencies measured in single-digit hertz. It is not a force machine, and displacement falls away quickly as frequency rises, because moving a tonne of assembly a millimetre at 200 Hz is an acceleration problem no screw-driven actuator solves.
Test programmes are written as environments rather than as machines, so one specification often contains clauses belonging to three different suppliers.
The 6-Class Test Input Crosswalk Read the clause, then choose the supplier
| The input your test needs | Typical range | The machine that produces it | Us? |
|---|---|---|---|
| Position and orientation reproduced over time, on a heavy assembly | 0 – 15 Hz | Servo electric motion platform | Yes — this is the machine on this page |
| Random, sine or swept vibration at small displacement | 5 – 3,000 Hz | Electrodynamic or servo-hydraulic shaker | No |
| Force or displacement along one axis at high rate | Load-controlled | Servo-hydraulic actuator or test frame | No |
| Shock, pyroshock or drop | Impulse | Shock machine or drop tower | No |
| Sustained acceleration beyond 1 g | Steady state | Centrifuge | No |
| Temperature, humidity or altitude | Environmental | Climatic chamber | No — though a platform can carry one |
The last row is worth a sentence. Platforms are regularly built to carry a chamber, a conditioning unit or a pre-soaked article, and that is a mechanical and thermal design question we are happy to take. What we do not supply is the chamber.
Ask which physics your clause describes before asking which supplier can quote it.
One specification with clauses from more than one of these rows? Send it and we will mark up which rows are ours →
Agreed before manufacturing
What a test host gets, and what happens at the workspace edge
Host link runs over TCP/UDP, serial or Modbus, with EtherCAT or CANopen below it. Command set, coordinate convention, pose order and units are fixed at the interface review and signed before anything is manufactured — a test rig is the wrong place to discover that two teams meant different things by “pitch positive”.
A commanded pose can be unreachable for reasons belonging to no single axis: a leg past its stroke, a joint past its angle, or a configuration where a small actuator motion produces a large platform one. Each pose and the path between poses is checked against all of those before anything moves, and one that fails is refused with a reason returned to the host — because a platform that quietly clips a commanded pose corrupts the data instead of stopping the run. The control system page carries the stack in full.
One safety boundary. The control interface can report status and request an orderly stop, but it does not replace emergency-stop circuits or the site safety chain. Those stay hard-wired and remain with whoever owns the installed system.
Integrating into a host that already drives other test hardware? Send its command convention and we will state what changes →
User wearing a VR headset in a boat motion simulator mounted on a motion platform
Written into the proposal, not discovered at handover
Where our scope ends and the test system begins
A test rig has more owners than a machine does, and most of the arguments happen at the seams. These three columns go into the proposal so the seams are agreed in writing while everyone still has options.
Ours
- Platform, controller, drives and electrical cabinet
- The agreed mechanical and electrical interfaces up to the mounting plane
- Profile conditioning, with every transformation documented
- Loaded platform FAT run against your conditioned profile
- Identification, acceptance and component records shipped with the machine
Yours, or your test integrator's
- The standard, the clause and the pass/fail criteria
- The article, its instrumentation and the data acquisition
- Everything above the mounting plane, fixture included
- Test-level analysis, the report and any certification
- Foundations, site power, permits and system-level safety
Signed by both
- The load case and the conditioned profile file
- The acceptance protocol and its tolerances
- The interface definition: command set, state packet, timing
- Safe-state, abort and reset behaviour
- The witness arrangement for the loaded FAT
What we do not do is certify anything. We are a machine builder, not an accredited test laboratory. We can state what the platform commanded, what it achieved and how repeatably, and that statement is instrumented and signed — but the judgement that an article passed a standard is made by your laboratory or your notified body, on their accreditation, not on ours.
Writing a tender that has to name responsibilities? Ask for this matrix as a blank annex →
From a test definition to an accepted bench
Five stages, and what each side hands over
A test platform is bought by a project, not by a purchaser, so the useful question is not lead time alone but what has to exist at each point and who produces it. Nothing here starts until the load case is agreed, because everything downstream is sized from it.
STAGE 01
Load case review
You send the profile sample, article and fixture masses, axes and acceptance target. We say whether it is our machine, and which of the four bench configurations it is.
Out: feasibility note
STAGE 02
Profile & envelope check
The recording is conditioned and run against the actuator envelope. Where it does not fit, you get the scaling factor or the axis that binds, not a refusal.
Out: conditioning report
STAGE 03
Interface freeze
Command set, state packet, coordinate convention, timing behaviour and safe state are written down and signed by both sides. After this the frame can be drawn.
Out: interface document
STAGE 04
Build & identification
Manufacture, assembly, and geometric identification on the as-built machine so the controller solves for the platform that exists rather than the nominal one.
Out: identification record
STAGE 05
Loaded FAT
Your conditioned profile run at the test load, repeated, with the run-to-run spread reported. Witnessed on site or by live video, your choice, before it is crated.
Out: acceptance record
Two of the five carry the schedule: Stage 03, where a project either gets an interface it can integrate against or inherits a commissioning problem, and Stage 05, where a supplier either runs your profile or runs its own.
Need this as a project plan for an internal approval? Ask for the five stages with your dates against them →
From the test side of the line
Reference reading for a test requirement
Four pieces test engineers ask for once a motion requirement is real: the axis decision, the trade-off, the interface and the delivered evidence.
Axis decision
3DOF or 6DOF, read from the requirement
What three axes cannot reproduce, what six cost you in moving mass, and how to tell which side of the line a recording falls on.
Trade-off
Payload, stroke and acceleration under a duty cycle
Why raising one of the three lowers the others, and why a continuous programme is sized against different numbers than a demonstration.
Interface
Command set, state packet and loop timing
What a real-time host needs from the platform, what the platform needs back, and what gets frozen before anything is manufactured.
Actuation
Why every bench on this page is servo electric
What hydraulic buys on a test rig, what it costs in maintenance, contamination and duty, and where the crossover actually sits.
Asked before a test method is fixed
Industrial testing motion platform FAQ
Can a motion platform replace a vibration shaker?
No, and the two are not close. A shaker produces small displacements at high frequency and is specified in g and in PSD; a motion platform produces large displacements at low frequency and is specified in position and attitude. A clause asking for random vibration to a spectrum in the hundreds or thousands of hertz is beyond any configuration of our machine. A clause asking for a recorded attitude or trajectory played back under a heavy article is exactly what it is for.
What frequency range can you actually reproduce, and how is that stated?
Test and measurement builds are issued with a working range of 0 to 15 Hz and a bandwidth of at least 2 Hz at −3 dB, both quoted at the rated load rather than empty. The figure that matters more than either is how much of your specific profile survives, so where the content is known we report tracking error against your file over the range it occupies instead of quoting a bandwidth in isolation.
Can we buy the platform only and integrate the test system ourselves?
Yes, and a large share of test work is bought that way. Configuration C2 exists for it: you get the platform, the controller and a written interface — command set, state packet, coordinate convention, timing behaviour and safe state — and you build the cell around it. Integrators can also take the control system alone and put it on their own mechanism. What does not change either way is that we still run a loaded acceptance before shipment, because a platform that has not been measured under load is not a test instrument yet.
Can you run our test to MIL-STD-810, IEC 60068 or ISO 16750?
We can build and instrument a platform that reproduces the motion a specific clause describes, where that motion is within the physics on this page. We cannot certify an article to any of those documents, because certification requires an accredited laboratory and we are a machine builder. In practice the useful conversation is about the clause rather than the document: send it, and we will tell you whether it describes our machine, a shaker, a shock machine, or a combination that needs more than one supplier.
Why do you keep asking about the fixture? We are buying a platform.
Because the fixture decides whether your article receives the motion you paid to generate. We control motion at the mounting plane; the fixture sits between that plane and your sensor and behaves as a filter or an amplifier depending on its stiffness and its first mode. We ask for its mass, CG and first natural frequency so we can tell you before the build whether the band you want to apply will survive the trip through it. The fixture stays yours either way.
Can I send road load or field-recorded data and have it played back exactly?
Send it, and expect the answer to include a list of what changed. Recorded data is resampled to the controller rate, filtered above the working range, corrected for integration drift and, where the trajectory leaves the workspace, scaled. Each transformation is reported as a number rather than applied quietly, so your report can state what was commanded and how it differs from the field recording. Exact playback of an unfiltered field recording is not something any platform does.
Why is your quotation higher than a supplier quoting the same payload and stroke?
Compare the conditions attached to the figures first. Ask whether the travel is single-axis or combined, whether the repeatability was measured with an instrument or read from encoders, whether it was measured at the test load, whether the quoted duty matches your cycle count, and whether acceptance runs your profile or a demonstration cycle. Those five differences account for most price gaps we see. Where a competing quotation genuinely covers the same conditions and costs less, we would rather hear it than guess.
Do you supply the data acquisition and write the test report?
No. Our instrumentation exists to prove the platform did what it was commanded to do, and that record ships with the machine. Measuring the article, running the analysis and issuing the report belong to the test owner or the test-system integrator, because those tasks require knowledge of what is being tested rather than of what is moving it. Where a customer wants a single supplier for both, we say so early rather than stretching the scope.
Commercial questions — lead time, payment terms, spares, on-site commissioning — are answered in the full FAQ.
Send the profile and the load case. We will tell you if it is our machine.
A sample of the data, the article and fixture masses, and the acceptance figure you have to meet are enough for a first pass. If the requirement belongs to a shaker, a test frame or a shock machine, that is what you will hear back — and hearing it in week one is worth more than a quotation.
What comes back first
- Whether the profile is inside our physics, naming the machine class if it is not
- A conditioning read of your sample file: sample rate, content, drift and the scaling it would need
- Which of the four bench configurations your programme is, and whether three axes or six suit the content
- A blank acceptance protocol and responsibility matrix, so both can be reviewed before an order