Motionwell Automation designs and builds machine tending automation in Singapore, the automated loading and unloading of production machines. Machining centres, lathes, presses, testers and moulding machines are all tended this way, and our delivered work sits at two ends of that range, metalcutting machine tools and test instruments. The reference at the heavy end is our own linear gantry robot series, where a steel single-beam gantry with 10 to 30 m of X-axis travel and dual Z-axis heads services more than ten CNC machines in parallel, one Z axis lifting finished parts out while the other puts blanks in. The series runs from a compact three-axis aluminium extrusion frame at 5 to 20 kg payload with pneumatic multi-finger grippers, through steel single-beam builds at 50 to 200 kg, to double-beam gantries carrying 200 to 1,000 kg and more across spans beyond 20 m. Machines are designed, assembled and tested at our Woodlands Link facility, and the company has delivered more than 150 special purpose machines since 2014 under ISO 9001:2015 and bizSAFE Level 3.
Where we stand, said plainly. We do not sell machine tools and we do not manufacture robots. We build the gantry, the tooling, the part presentation, the machine interface and the control system that ties them together, and where a standard cell from your machine tool builder covers the job at a lower price we say so at concept review. Our mechanical designers have direct experience operating CNC machines, which keeps out designs that look good on screen and are impractical to manufacture. The delivered gantry work sits in context on our warehouse and intralogistics page; if you already have a machine list and a cycle time, skip ahead and talk to an engineer.
Which Robot Type Should Tend Your Machines?
Three families cover almost every machine tending cell you will be quoted, and three measurements settle the choice before anyone opens CAD: how far the robot must travel, how much floor you can give it, and how many machines it has to serve.
| Selection factor | Linear gantry | Articulated six-axis arm | Collaborative arm |
|---|---|---|---|
| Machines served per unit | Ten or more on a straight row | One or two within reach, more on a seventh-axis track | Usually one |
| Travel | 1 to 5 m compact, 10 to 30 m single beam, beyond 20 m double beam | Reach-limited unless extended on a linear track | Short, sized around one machine |
| Floor space | Travels overhead, so the floor stays clear for operators and pallet trucks | Its own cell area plus access | Small footprint, shared with the operator |
| Payload | 5 to 20 kg compact, 50 to 200 kg single beam, 200 to 1,000 kg and above | Falls off sharply at full extension | Rated payload minus the gripper, less again at reach |
| Orientation freedom | Three axes, so the part enters as the gantry presents it | Full six-axis, which suits an angled chuck or tilted fixture | Full six-axis, at low speed |
| Where it fails | A scattered layout, or a part that must enter at an angle | Long rows, because reach is the wrong tool for distance | Rate, and payload once the gripper is counted |
The first row decides most projects. An arm is bought for orientation freedom and pays for it in reach; a gantry is bought for distance and pays for it in orientation. If your parts drop vertically into a chuck and your machines sit in a line, you are paying for articulation you will never use. If the part enters an angled fixture, no amount of gantry travel substitutes for a wrist.
Collaborative arms earn their place where the duty is one machine, the part is light and a person is genuinely in the loop. Our delivered cobot tending is test and laboratory work rather than metalcutting: a cobot loads universal testing machines in the QA laboratory automation programme re-ordered in four consecutive years, and another loads specimens into the enclosure on our cleanroom automated test equipment, holding ±0.05 mm at the fixture interface. A force-limited arm is slow by design, and that slowness compounds across every machine it reaches.
Why Does One Long Gantry Beat Several Arms on a Row of Machines?
Because distance is cheap on a beam and expensive on a wrist, and because the machines already sit in a line for reasons that have nothing to do with robots.
A gantry extends along X in modules. The X axis runs on rack and pinion and the beam is built in sections, so two more machines on the row means more beam and rail rather than another robot, another controller, another tooling set and another fenced footprint. An arm does not scale that way: reach cannot be extended without changing arm class, and payload falls off at full extension exactly where you need it. Travelling overhead also leaves the floor clear for the operator, the pallet truck and the swarf bin, and that aisle is often worth more than the difference in robot price.
We put the cost argument as an engineering trade rather than a discount. For a row of machine tools served by one overhead gantry, capital and cell area come out materially below deploying several six-axis arms for the same job, because you buy one motion system instead of several and fence one perimeter instead of several. That reverses the moment the machines stop being in a row: a scattered layout, machines on both sides of an aisle, or a part that must be reoriented between operations all push the answer back towards arms.
| Series | Payload | X-axis travel | Traverse speed | Repeatability | Z axes | Where it fits |
|---|---|---|---|---|---|---|
| Compact three-axis, aluminium extrusion frame | 5 to 20 kg | 1 to 5 m | 1.5 m/s | ±0.1 mm | 1 | Light parts, small machines, transfer between assembly stations |
| Steel single beam | 50 to 200 kg | 10 to 30 m | 1.2 m/s | ±0.2 mm | 1 or 2 | A row of CNC machines, ten or more tools from one gantry |
| Double beam | 200 to 1,000 kg and above | Beyond 20 m, to 50 m on the largest builds | 1.0 m/s | ±0.3 mm | 1 or 2 | Heavy workpieces, wide spans, with maintenance walkway and railings |
Two design points inside that table matter more than the numbers.
Dual Z axes. The single-beam and double-beam builds can carry two independent Z columns, each with its own gripper pair. One column lifts the finished part out while the other lowers the next blank in, so load and unload happen in one visit instead of two. Across a row of machines that halves the number of visits, and it is the visits, not the reach, that limit how many machines one gantry can hold.
Repeatability is not placement accuracy. A gantry repeating to ±0.2 mm tells you nothing about whether the part seats in the chuck. Seating is decided by how well the blank was located before the pick, by gripper repeatability, by chip on the locating face and by fixture wear. Where a blank cannot be located mechanically, a camera measures its actual position before each pick and the controller corrects the move, the approach set out on our machine vision inspection page.
What Does the Machine Tool Have to Tell the Robot Before It Can Be Loaded?
The signal list below is the foundation the rest of the cell is designed on, because a robot cannot tend a machine that cannot talk to it. The machine has to say that it has finished, that the door is open and that the workholding has let go, and it has to hold still while steel is in its work envelope.
| Signal | Direction | What it means | What goes wrong without it |
|---|---|---|---|
| Cycle complete | Machine to robot | The programme has ended and the spindle has stopped | The robot enters on a timer and meets a moving tool |
| Door open, confirmed | Machine to robot | The door has physically reached the open position | The gripper hits a door that stuck halfway |
| Chuck or vice open, confirmed | Machine to robot | Workholding has released, sensed rather than commanded | The robot pulls against a clamped part and drops it or wrecks the tooling |
| Part present or clamped | Machine to robot | A part is in the fixture and gripped | The machine cuts air, or the robot loads onto a part still there |
| Robot in envelope | Robot to machine | Hold the machine: no motion, no door close | The door closes on the arm |
| Cycle start | Robot to machine | The envelope is clear, begin the programme | Someone has to press a button, so nothing runs unattended |
A fault and mode signal in both directions belongs on that list too, so the cell stops feeding a machine that has already stopped. Newer machine tools generally offer a robot interface option, providing most of it as a defined connector and a set of auxiliary functions, usually programmed as spare M-functions in the part programme. On an older machine that option often does not exist and the interface has to be added. Do that through the machine tool builder or their agent wherever possible. The door interlock and the chuck clamp are safety functions on that machine, so taking control of them moves responsibility onto whoever did the work, and a modification substantial enough can put that party into the manufacturer’s position for the modified machine. Our machine safety and CE marking page works that through.
The control-side half is routine here. Control system modernisation of legacy machines, replacing ageing PLCs, servo drives and VFDs on Allen-Bradley, Siemens, Omron, Mitsubishi and Beckhoff platforms, is our largest line of work this year. Where a controller is too old to extend safely we quote the replacement instead of bolting a second one alongside it.
How Should Parts Be Presented to the Robot?
Part presentation decides how much the robot has to work out for itself before it picks, and that is what sets the sensing, the tooling and the cycle time around it.
| Presentation | How the robot finds the part | Suits | What it costs you |
|---|---|---|---|
| Bar or billet magazine | Mechanical, one part at a time at a fixed pick point | Turned parts, repetitive blanks of one size | A magazine per blank size, and a jam clears by hand |
| Fixtured tray or pallet | Taught grid, indexed by row and column | Machined blanks, medium mix, batch running | Trays are a real cost per part family, and someone loads them |
| Conveyor with stop and locate | Driven against a stop and squared before the pick | Continuous feed from an upstream process | Needs presence sensing and a squaring device, not a timer |
| Random bin | Nothing is known until a 3D sensor measures it | High-mix unsorted castings and forgings | A different project, see below |
The first three share one property: position is engineered before the robot arrives, by a fixture or by a locating motion. That is why they are predictable to quote, and why position is confirmed with SICK and ifm sensing rather than assumed from a timer.
Tooling follows the part rather than the robot. The compact three-axis frames in our linear gantry series run pneumatic multi-finger grippers, and the truss manipulators run pneumatic grippers sized to the load; where a second tool has to fit without re-machining the Z column we design a quick-change interface in. For mixed part families, changeover frequency decides between one combination tool that compromises on every part and a tool rack: several changes a shift argues for the combination tool, weekly changes argue for the rack. Grippers come from Schunk and DH Robotics, pneumatics from SMC and Festo, and the application tooling is designed in house.
Why Is Bin Picking a Different Project?
Because everything above depends on the part’s position being known before the robot moves, and a random bin removes that assumption. From a bin the robot has to perceive the scene, choose a part that is on top and reachable, plan a collision-free approach into a steel container, grip a part whose orientation it did not choose, then place it in a known orientation, which usually means a regrip station. That last step catches people out: picking successfully from the bin does not put the part in the chuck the right way round.
On our own record: Motionwell’s delivered picking is from fixtured, taught or vision-corrected positions, and we have not delivered a random bin picking cell. If your parts genuinely arrive heaped, treat it as its own project with a proving trial on your actual parts, and check first whether tray or magazine presentation upstream is cheaper than perception downstream. Very often it is.
What Do Swarf, Chips and Coolant Do to a Tending Cell?
They are the environmental problem that decides whether the cell still runs in year three, and they are consistently under-scoped. Four places take the damage.
The mechanism. Linear guides, ball screws and rack teeth collect chips, and coolant mist strips grease. This is one of the arguments for overhead travel: the structure sits above the mess rather than in it. Where the environment is genuinely dirty we use V-groove roller guides rather than recirculating ball guides, which is what our dual-rail truss manipulators use for dusty and harsh environments, because a roller on a V edge sheds debris a ball guide would ingest.
The cabling. Every gantry axis carries its cables in a cable carrier, and a carrier filling with hot chips becomes an abrasion problem long before it becomes a jam. Route it away from the chip conveyor discharge and specify continuous-flex cable rated for the travel.
The gripper and the part. A chip on a locating face is a common cause of a part that will not seat, and it is invisible on a drawing. The fixes have to be designed in: an air blast at the chuck and at the gripper before every load, a locating scheme that seats on features a chip cannot sit on, and a clamp confirmation that measures rather than assumes. Coolant carried out on a wet part is also a floor hazard and a gripper slip risk, which is why a drain and dwell position on the way out earns its cycle seconds.
The sensors. Optical devices under a coolant film report whatever they like, so inductive sensing is the safer choice on metal parts in a wet environment.
How Does the Cycle-Time Arithmetic Actually Work?
Machine tending is the one application where an idle robot is the correct outcome, and misreading that is a common reason a sound project gets rejected.
The machine cycle sets the pace and the robot’s job is to finish inside that window. Per machine, tending time is the sum of travel to the machine, wait for door open, reach in, unclamp handshake, remove the finished part, place the blank, clamp handshake, withdraw, door close and cycle start. On a dual Z-axis head the remove and the place happen in one visit rather than two, which is where that design earns its cost. How many machines a gantry can hold then falls out of one line:
Machines served ≈ shortest machine cycle ÷ (tending time per machine + travel between machines)
Run it with your own numbers before anyone quotes hardware. It explains the two results that surprise buyers most. A long machining cycle leaves the robot idle most of the time, which is not waste: it is the headroom that lets one gantry cover ten machines instead of one. And a row of fast machines is harder than a row of slow ones, because tending time barely changes while the window shrinks. A mixed row bottlenecks on whichever tool finishes fastest, and the fix is usually sequencing or a second Z head rather than a faster gantry.
What Does Unattended Running Actually Demand?
Running with nobody watching is a different specification, not a longer shift. Five things have to be true, and each is a cost.
Enough material in, and enough space for parts out. The cell runs until the input empties or the output fills, so magazine depth and finished-part buffer, measured in minutes at your cycle rate, are specifications worth writing down. This is what quietly decides how many unattended hours you get.
Every abnormal condition detected, not only the ones you thought of. Part present, part gripped, chuck clamped, door confirmed, tool life, coolant level, chip conveyor running, air pressure. The cell fails safe by stopping, and it can only stop on a condition it can sense.
A defined behaviour on every fault. Stop and hold is the right default. Where a fault is transient a bounded retry beats an immediate stop, and it must still be logged. We use that pattern in the QA lab cell, where each positioning layer retries its tolerance check up to three times before escalating with the failure mode recorded, on the reasoning that a fault recovered silently is worse than a stoppage.
A record of what happened while nobody was there. Counts and timestamped alarms, so the morning shift can tell whether the cell stopped at 22:15 or 05:50. That difference is a night of spindle hours.
Guarding that holds with nobody in the room. Unattended running does not reduce the safety scope; it changes who carries the residual risk, because whoever opens the guard next arrives at a cell in an unknown state. Restart should need a deliberate reset with a view of the whole cell, never a power cycle. We have delivered cobot-assisted loading for unattended operation on the 5-axis CNC shot peening machine built for turbine blade surface treatment.
What Safety Scope Applies to a Machine Tending Cell in Singapore?
Start from an ISO 12100 risk assessment and let it drive the design rather than the reverse. A tending cell carries a specific hazard mix: a robot moving over long travel, a machine tool with its own energy sources, a door that closes, workholding that clamps with real force, and an operator who reloads magazines and clears chips while the cell is live.
ISO 10218-1 covers the robot and ISO 10218-2 the integrated system, which is the part a machine builder owns. Each safety function then carries a required performance level under ISO 13849-1, calculated and validated rather than asserted. The 2025 edition of ISO 10218-1 is current and the first substantive revision since 2011; it references the 2023 edition of ISO 13849-1, so a design documented against the older edition will need its performance level calculations restated when the cell is reassessed.
Three points specific to tending are worth settling before layout is fixed. The boundary between two safety systems: the machine tool has its own door interlock and stop category and the cell has its own, so somebody must own the boundary and prove the robot cannot enter while the spindle can turn. Stopping performance after a retrofit: measure real stopping time, then confirm the scanner distances calculated to ISO 13855 still hold, because a new servo drive on old mechanics changes run-down time in ways no datasheet predicts. Local requirements: Singapore does not require CE marking, but the Workplace Safety and Health Act requires dangerous parts of machinery to be securely fenced, and machines incorporating lifting equipment must be certified by an approved authorised examiner before use.
We deliver the physical scope that follows: guard fencing and interlocked access doors, safety laser scanners, safety-rated stop circuits, LVD and CE testing, and MOM lifting certification. Where a collaborative arm is used instead, force limiting has to be proven rather than assumed, and it stops being collaborative the moment a sharp gripper or a heavy blank is fitted, as our note on collaborative robot safety standards sets out.
What Drives Cost and Lead Time on a Machine Tending System?
We do not publish prices, because two quotations for what looks like the same machine tending robot in Singapore can differ by a wide margin on decisions taken before hardware is ordered. Two items belong in the quotation and are easy to leave out of one. Retrofit archaeology, establishing what the existing circuits actually do before anything is replaced, is regularly the largest single line in a retrofit safety scope. And rate has to be timed on your machines rather than taken from a datasheet, because handshake waits on a legacy interface are frequently a larger term than the robot move itself, for the reasons in our guide to what OEE actually measures. Beyond those, here is what moves the number.
| Cost driver | Why it moves the number |
|---|---|
| Machine count and row length | Sets beam length, rail, cable carrier and drive sizing, and it is the cheapest thing to get right at concept |
| Single or dual Z axis | A second Z column is a second set of drives and tooling, bought back in trips saved per machine |
| Part family count | Each family is a gripper, a fixture, a tray and a recipe; format count costs more than it looks |
| Part presentation | A magazine or fixtured tray is engineering you can price; perception into a random bin is a project |
| Machine interface state | A robot-ready machine tool is a connector; a legacy row is an interface per machine |
| Safety scope | Perimeter length, access points, scanner zones, and stopping performance measurement after retrofit |
Lead time runs 16 to 24 weeks from concept approval to factory acceptance testing on a standard build, and 24 to 32 weeks where cleanroom compatibility or full validation documentation applies. Design, fabrication, assembly and testing happen at Woodlands Link with an in-house design team of eight, which is why a Singapore buyer attends the factory acceptance test instead of flying to it. That matters on tending work, because the adjustments that decide whether a cell runs are gripper and fixture details found on your own parts. How a build of this shape is scoped and commissioned is covered in our special purpose machine design guide.
What Will We Not Take On?
We do not supply machine tools and we do not rebuild them. We do not manufacture robots; we design and build the gantry and the tooling and integrate arms across FANUC, ABB, KUKA, Yaskawa, Yamaha and JAKA platforms, with integration experience on customer-supplied Universal Robots cells across three projects. We have not delivered a random bin picking cell, and we will say so rather than learn on your parts. We do not take on production welding cells. We do not issue CE certificates and we are not a notified body; we build to a specification and support your conformity work. And where a standard tending cell from your machine tool builder covers the job at a lower price, that is the useful answer. Downstream, the same handling engineering continues into automated palletizing for finished parts.
Which standard editions apply right now?
The editions below are the ones we design and document against on current projects. We check them on the date shown rather than assuming last year's edition still holds.
| Standard | Current edition | What it means for your machine |
|---|---|---|
| ISO 10218-1 — Robotics, safety requirements, Part 1: industrial robots | ISO 10218-1:2025 | Published February 2025, the third edition and the first substantive revision since 2011. It adds robot classifications with matching functional safety requirements, safety-related cybersecurity requirements, and end-effector guidance. Most of ISO/TS 15066:2016 on collaborative operation moved into Part 2. |
| ISO 13849-1 — Safety of machinery, safety-related parts of control systems | ISO 13849-1:2023 | The 2023 edition is the version referenced by ISO 10218-1:2025 for robot control system safety functions. Designs still documented against the 2015 edition will need their PL calculations restated when the machine is re-assessed. |
Frequently Asked Questions
How many machine tools can one gantry robot serve?
More than ten on the systems we have delivered, using a steel single-beam gantry with 10 to 30 m of X-axis travel and a dual Z-axis head. The real limit is arithmetic rather than mechanics: one gantry serves as many machines as it can complete a full tending round trip for inside the shortest machine cycle on the row. Divide the shortest cycle by the tending time per machine plus the travel between machines, and that is your ceiling. Very short cycles, or a mixed row where one tool finishes far faster than the rest, bring the number down quickly.
Do I need to modify my CNC machine to add a robot?
Almost always, if the machine is more than a few years old. The robot needs the machine to tell it that the cycle is finished, that the door is open, and that the chuck or vice has actually released, and it needs to hold the machine while a tool is inside the work envelope. Newer machine tools generally offer a robot interface option providing those signals. Legacy machines usually need an interface added, and the safe route is through the machine builder or their agent, because the door interlock and the clamp are safety functions on a machine that was signed off without you.
Is machine tending worth automating if the robot waits most of the cycle?
Yes, and robot idle time is the wrong metric. A tending robot is bought to take an operator off a repetitive load and unload task and to keep the machine cutting through breaks, shift changes and the night. On a long machining cycle the robot is idle by design, and that idle time is exactly the headroom that lets one gantry cover a row of machines instead of one. Judge the investment on machine utilisation and on spindle hours run unattended, not on how busy the robot looks.