Robotic Machine Tending

I'll tell you something that surprises people who don't live on the shop floor: the biggest bottleneck in most machining operations isn't the machine. It's the person loading and unloading it. Your CNC machining center cuts metal at spindle speeds that would've been science fiction 20 years ago—but the spindle sits idle every time an operator opens the door, reaches in, swaps a part, closes the door, and hits cycle start. On a typical 90-second machining cycle, that load/unload sequence eats 25–40 seconds. And when the operator takes a break, goes to lunch, or calls in sick, the spindle doesn't turn at all.

At AMD Machines, we've built robotic machine tending systems for over two decades—hundreds of cells across automotive, aerospace, medical device, and heavy equipment manufacturing. We've learned what works, what breaks, and what separates a machine tending cell that actually runs lights-out from one that needs babysitting. This page covers everything we know.

The Real Economics of Machine Tending

Before we talk technology, let's talk numbers—because the math on machine tending is some of the most compelling in all of manufacturing automation.

Most machine shops run their CNCs at 40–55% spindle utilization across a standard two-shift operation. That means your $300,000 machining center is cutting metal less than half the time you're paying for the building, the electricity, and the machine's depreciation. The rest is operator load/unload time, breaks, shift changes, setup, and unplanned downtime.

A robotic tending cell pushes spindle utilization to 85–92%. Here's where that improvement comes from:

• Load/unload time drops from 25–40 seconds to 6–12 seconds. The robot doesn't fumble with parts, doesn't need to find the wrench for the vise, and doesn't take a phone call mid-cycle.
• Breaks and shift changes disappear. The robot runs through lunch, through breaks, through shift changes. That's 1.5–2 hours of recovered spindle time per shift.
• Third shift and weekends become productive. Lights-out running adds 40–60 hours of weekly spindle time that you simply can't get with manual labor—nobody wants to run a CNC alone on the night shift, and even if they did, the labor cost would eat your margins.

On a two-shift operation machining parts with a 90-second cycle, going from 50% to 88% spindle utilization means producing 580 more parts per day. At even modest per-part margins, that's revenue you're currently leaving on the table.

Machine Tending Cell Configurations

We design every tending cell around three factors: your cycle time, your part mix, and your floor space. There's no one-size-fits-all configuration, but after hundreds of installations, we've found that nearly every application fits one of these architectures.

Single-Machine Cells

One robot, one machine. This is the most common configuration and the right starting point for most shops.

A FANUC M-10iD/12 or LR Mate 200iD sits in front of the CNC with an infeed conveyor on one side and an outfeed on the other. The robot picks a raw blank from the infeed, opens the machine door (or signals the automatic door), loads the part into the chuck or vise, closes the door, starts the cycle, and waits. When the cycle completes, it unloads the finished part, blows chips off the fixture if needed, and loads the next blank.

Cell footprint is typically 2.5 m x 3 m—compact enough to fit beside the machine without major floor plan changes. We've installed single-machine cells in shops with 3-meter aisle widths by using a FANUC CRX-10iA/L collaborative robot that doesn't require full safety fencing, just area scanners and speed-limited zones.

Best for: High-value parts with cycle times over 60 seconds, shops adding their first automation, and dedicated production of a single part family.

Multi-Machine Cells

One robot tending two, three, or even four machines. This is where the economics get really interesting.

If your machining cycle is long enough—typically 3+ minutes per part—the robot has idle time between load/unload operations on a single machine. A multi-machine cell fills that idle time by having the robot service additional machines in sequence. We've built cells where a single FANUC M-20iD/25 robot tends three Mazak HCN-5000 horizontal machining centers, keeping all three machines above 90% spindle utilization on a 4.5-minute cycle.

The key to making multi-machine cells work is cycle time balancing. We use simulation software during the design phase to model robot travel times, machine cycle times, and part staging sequences. If the cycle times don't balance—if the robot can't get back to Machine 1 before it finishes its cycle—you get machine idle time that defeats the purpose. We build in time buffers and optimize the robot path to ensure every machine gets serviced within its window.

For multi-machine cells, we typically use FANUC M-10iD or M-20iD robots mounted on a riser for extended reach, or ABB IRB 2600 robots when the machine layout requires longer reach to access multiple machine doors.

Best for: Parts with cycle times over 3 minutes, shops looking to maximize robot ROI, and production environments running families of similar parts across multiple machines.

Linear Rail Systems

When you need to tend a line of five or more machines, putting the robot on a linear rail (seventh axis) is more economical than buying a robot for every two or three machines. The robot travels along the rail, stopping at each machine to perform load/unload operations.

We integrate FANUC robots on Güdel or FANUC-native rail systems. The rail adds a seventh servo axis that the robot controller manages seamlessly—the robot treats the rail position as just another joint in its motion plan. Travel speed along the rail is typically 1–2 m/s, so the robot can move between machines spaced 3 meters apart in under 3 seconds.

On a recent heavy equipment project, we installed a FANUC M-710iC/50 on a 12-meter Güdel rail tending six Okuma MULTUS CNC lathes. The system runs three shifts, producing large-diameter hydraulic cylinder components with a 7.5-minute machining cycle. One robot, one operator per shift for material staging, six machines running at 91% utilization.

Best for: Five or more machines in a line, parts with longer cycle times (5+ minutes), and facilities with the floor space for a linear cell layout.

Rotary Indexing Machine Tending

For very short cycle times—under 30 seconds—the robot can't keep up with a single high-speed machine using conventional pick-and-place. In these applications, we integrate the machine tending function into a rotary indexing system with dedicated load and unload stations. The indexer moves parts between stations on a continuous dial, and the machining operation happens at one or more stations in the rotation sequence.

Best for: High-volume parts with cycle times under 30 seconds, operations that combine machining with other processes (inspection, marking, assembly).

Part Handling and Staging

How parts get into and out of the cell matters as much as the robot itself. We've seen too many tending cells where the machine runs beautifully but the operator spends their entire shift feeding parts into a conveyor. That's not automation—it's just moving the bottleneck.

Conveyor-Fed Systems

Infeed and outfeed conveyors provide continuous part flow. The operator loads raw blanks onto the infeed conveyor and removes finished parts from the outfeed. With properly sized conveyors, one operator can stage enough parts for 30–60 minutes of unattended running, then walk away to tend other equipment.

We use Dorner and mk North America conveyors with part nesting fixtures for oriented feeding, or flat belt conveyors with vision-guided picking when parts can arrive in random orientation.

Drawer and Pallet Systems

For maximum unattended run time, drawer or pallet systems let you stage hours worth of parts. The operator loads a drawer of raw blanks and removes a drawer of finished parts without stopping the cell. We use Erowa, System 3R, and Lang Technovation pallet systems that interface directly with the CNC's pallet receiver, and custom drawer systems that hold 4–8 hours of production inventory.

On a lights-out aerospace cell we built, four Erowa pallets hold 96 titanium blanks each—enough for 16 hours of unattended production through the night shift and into the next morning. The operator loads pallets at 6 AM and the cell runs until 10 PM without intervention.

Bin Picking

When parts arrive in bulk—castings dumped in a bin, forgings stacked in a tote—robotic bin picking with 3D vision eliminates the need for organized part presentation. We integrate Keyence or Photoneo 3D vision sensors with FANUC's iRVision system to locate and pick parts from random orientations in bins or totes.

Bin picking works well for robust parts that tolerate gripper contact—castings, forgings, stampings—but isn't ideal for finished surfaces that scratch easily. For those parts, we use organized staging.

Integration with CNC Controls

The robot and the CNC need to talk to each other reliably. Door open, door closed, chuck clamped, chuck unclamped, cycle start, cycle complete—these handshake signals are safety-critical. If the robot reaches into the machine while the spindle is turning, you don't get a second chance.

We integrate with all major CNC controls:

• FANUC CNC — Direct interface via FANUC's robot/CNC connect protocol over EtherNet/IP. Fastest integration with FANUC robots since both controllers speak the same language.
• Siemens SINUMERIK — PROFINET communication for I/O handshaking and cycle management. We use Siemens safety I/O modules for door interlock and chuck confirmation signals.
• Mazak SmoothCNC — Ethernet communication with Mazak's automation interface. We've done dozens of Mazak integrations, especially on INTEGREX multi-tasking machines.
• Okuma OSP — Ethernet/IP integration with Okuma's robot interface. Okuma's ESAS (Easy Setup Assist System) simplifies the initial handshake configuration.
• Haas — Serial or Ethernet communication via Haas's automation interface. Haas machines are the most common CNC platform in North American job shops, and we've standardized our tending cell designs for the most popular Haas models.

Every integration includes a safety interlock system that prevents robot motion inside the machine enclosure unless the spindle is stopped, the chuck or fixture is in a confirmed state, and the machine door is fully open. We use safety-rated I/O (Category 3 / PLd minimum per ISO 13849) for all interlock signals.

Quality Integration in the Tending Cell

A machine tending cell that only loads and unloads is leaving value on the table. The robot is already handling every part—so we use that handling time to integrate quality operations that eliminate downstream inspection bottlenecks.

In-Cell Gauging

We integrate contact gauging stations (Renishaw Equator, Marposs, Keyence GT2 series) inside the cell. The robot places the machined part on the gauge fixture between the machine unload and the outfeed conveyor. Critical dimensions are checked on every part or at a statistical sampling rate. If a dimension trends out of tolerance, the system automatically adjusts CNC tool offsets to compensate for tool wear—closed-loop machining without operator intervention.

On a medical device machining cell for orthopedic implants, we integrated a Renishaw Equator gauging system that checks 12 critical dimensions on every titanium knee component. The gauging cycle takes 8 seconds—hidden inside the 4.5-minute machining cycle—and the system has maintained Cpk > 1.67 on all critical features for over 18 months of production.

Vision Inspection

Machine vision cameras (Keyence CV-X, Cognex In-Sight 2800) inspect parts for surface defects, burrs, missing features, and wrong-part detection. We mount cameras at the robot's unload position so inspection happens during the part transfer—no additional cycle time.

Part Marking

Laser marking or dot peen marking inside the cell provides traceability for every part. The robot places the part at the marking station, the mark is applied, and a vision system verifies the mark readability before the part exits the cell. Full traceability from raw material to finished part, with no manual data entry.

Real-World Application Examples

Automotive: Transmission Gear Machining Cell

A Tier 1 automotive supplier machining transmission gears was struggling with capacity. Their six Mazak INTEGREX 200-IV multi-tasking machines were running two shifts with manual tending, producing 1,800 gears per day at 52% spindle utilization. Adding a third shift wasn't viable—they couldn't find machinists willing to work nights, and the labor cost destroyed margins on the contract.

We built three dual-machine cells, each with a FANUC M-20iD/25 robot tending two INTEGREX machines. Infeed drawers hold 120 gear blanks (4 hours of production), and outfeed drawers collect finished gears. A Marposs gauging station in each cell checks gear OD, bore diameter, and face runout on every part, with automatic tool offset correction fed back to the CNC.

Results: Spindle utilization increased from 52% to 89%. Daily output jumped from 1,800 to 3,100 gears—a 72% increase—without adding any headcount. Two operators per shift manage all three cells (staging material and monitoring). The system runs lights-out from 11 PM to 6 AM, producing 680 gears overnight. Payback was 14 months. Cpk on critical bore diameter improved from 1.35 (manual) to 1.72 (automated) because the gauging system catches tool wear drift that operators missed.

Aerospace: Titanium Bracket Machining

An aerospace manufacturer producing titanium brackets for a commercial aircraft program needed to increase output without adding machines. Their five-axis Okuma MU-5000V machining centers had a 22-minute cycle per bracket, and operators were managing to load/unload in about 45 seconds—but the machines sat idle during breaks, shift changes, and the entire third shift.

We installed a FANUC M-710iC/50 robot on a 9-meter Güdel linear rail, tending four Okuma machines. Erowa pallet systems provide 12 hours of unattended raw material staging. The robot handles titanium blanks weighing up to 8 kg, loading each into a hydraulic fixture with ±0.05 mm repeatability. A Renishaw Equator gauges three critical dimensions on every part.

Results: Spindle utilization went from 61% to 93%. Weekly output increased by 52% with the same four machines. The cell runs lights-out 5 nights per week, adding 50 hours of production that simply didn't exist before. The customer delayed a planned $1.2M capital expenditure for a fifth machining center because the tending cell gave them the capacity they needed from the existing four.

Medical: Orthopedic Implant Micro-Machining

A medical device company machining titanium and cobalt-chrome orthopedic implants needed automation that met FDA quality requirements while handling parts too small for standard grippers. Each implant weighs 12–45 grams, requires 100% dimensional inspection, and demands full lot traceability per 21 CFR Part 820.

We designed a cell around a FANUC LR Mate 200iD/7L robot with custom vacuum grippers that handle the parts by their non-critical surfaces. A Keyence GT2 contact probe gauges five critical dimensions on every part in an 11-second gauging cycle. Laser marking provides a unique device identifier (UDI) per FDA requirements, and a Cognex DataMan 370 verifies the mark readability before the part exits the cell.

Results: 100% in-process dimensional inspection (vs. statistical sampling before), Cpk > 2.0 on all critical dimensions, full electronic batch records meeting 21 CFR Part 820. Scrap from machining defects dropped from 2.1% to 0.3%. The cell runs unattended for 6-hour stretches between material staging, producing 280 implants per day vs. 190 with manual tending—a 47% throughput increase.

The ROI of Machine Tending Automation

The payback math for machine tending is straightforward, and it stacks up fast.

Increased spindle time is the primary driver. Going from 50% to 88% utilization on a machining center producing parts worth $15 each in margin adds roughly $200,000–$350,000 in annual revenue per machine—without buying a new machine. If you're capacity-constrained, this is the fastest path to more output.

Labor savings are the second lever. A machine tending cell doesn't eliminate operators—it changes what they do. Instead of standing in front of one machine all day, one operator manages 3–6 machines, staging material and monitoring cell status. On a two-shift operation, that's a reduction of 2–4 operators at a fully burdened cost of $55,000–$70,000 each, saving $110,000–$280,000/year.

Quality improvement is the third lever. In-cell gauging with automatic tool offset correction catches dimensional drift before it produces scrap. Customers consistently report 40–60% reductions in machining scrap rates after implementing robotic tending with closed-loop gauging. On a 100,000-part annual volume with a $25 scrap cost per part, reducing scrap from 3% to 1% saves $50,000/year.

Capital avoidance is the lever nobody talks about until it happens. If you can get 88% utilization from four machines instead of buying a fifth, you've avoided a $300,000–$500,000 machine purchase plus $100,000+ in installation and tooling.

A typical single-machine CNC tending cell with a FANUC robot, conveyor staging, and basic gauging costs $150,000–$250,000. Multi-machine cells with rail systems and advanced gauging run $300,000–$500,000. We see full payback in 8–16 months on two-shift operations, and customers running lights-out see even faster returns.

Common Challenges and How We Solve Them

"We run dozens of different part numbers." Flexible fixturing and gripper changeout solve this. We design quick-change gripper systems (typically Schunk SWS or ATI tool changers) that let the robot swap grippers in 3 seconds. For vise-based CNC setups, we use hydraulic fixtures with adjustable jaws that accommodate a family of parts without manual changeover. The robot retrieves the correct gripper from a tool stand based on the production schedule downloaded from your MES or ERP system.

"Our parts have chips and coolant on them." Chip management is one of the most overlooked aspects of machine tending. We integrate air blow-off nozzles in the robot's path—the robot holds the part in front of an air knife for 2–3 seconds to clear chips and coolant before placing it on the gauge or outfeed. For stubborn chips in internal features, we add a chip wash station with directed coolant jets. The machine's chip conveyor and coolant filtration system need to keep up with continuous production—we review your chip management setup during the design phase and recommend upgrades if needed.

"We can't justify automation for short production runs." This is where multi-machine cells and flexible fixturing change the equation. If you're running 50-part batches but you have 20 different part numbers flowing through the same machines, the robot doesn't care about batch size—it cares about cycle time balance. We program all your part numbers, set up quick-change grippers for each part family, and let the cell schedule pull the right program automatically. Changeover time drops from 30–45 minutes (manual) to under 60 seconds (automated gripper swap plus program recall).

"Our machines are old and don't have automation interfaces." We retrofit older machines with the I/O interfaces needed for robotic tending. A CNC machine doesn't need to be new to be automated—it needs a door actuator (we install pneumatic or servo door openers), a chuck/vise confirmation signal, a cycle start input, and a cycle complete output. We've retrofitted machines from the 1990s with modern automation interfaces. If the machine makes good parts, we can automate it.

"Lights-out scares us—what if something goes wrong at 2 AM?" Remote monitoring. Every cell we build includes Ethernet connectivity to your plant network, and we integrate with monitoring platforms that send alerts via text and email when the cell stops, a fault occurs, or material runs low. Some customers use cameras inside the cell for visual verification. We also build in automatic recovery logic—if the robot drops a part or a gauge cycle fails, the system retries before faulting. On most cells, over 95% of minor faults resolve automatically without human intervention.

Frequently Asked Questions

What robots do you use for machine tending?

We primarily integrate FANUC robots for machine tending—the M-10iD, M-20iD, and LR Mate 200iD are our workhorse models covering payloads from 7 kg to 25 kg. For heavier parts (castings, forgings), we use FANUC M-710iC/50 (50 kg payload) or ABB IRB 4600 robots. For shops that want collaborative operation without full safety fencing, we integrate FANUC CRX-10iA/L or CRX-25iA cobots with area scanners. Robot selection depends on part weight, reach requirements, and cycle time targets—we model it all in simulation before committing to hardware.

How long does it take to program a new part number?

For a standard CNC tending operation with vise fixturing, teaching a new part takes 1–2 hours, including gripper setup, pick/place positions, and machine interface verification. If you're using offline programming (we offer FANUC ROBOGUIDE setup as part of our robot programming services), you can program parts without taking the cell offline. For high-mix shops, we create parametric programs where the operator enters part dimensions at the HMI and the robot calculates pick positions, place positions, and gripper jaw settings automatically.

Can one robot tend different types of machines?

Yes. We've built cells where a single robot tends a CNC lathe, a CNC mill, and a parts washer in sequence—the part starts as a rough turning, moves to the mill for secondary features, gets washed, and exits the cell as a finished part. The robot uses different grippers for each machine (raw blank vs. semi-finished vs. finished), swapping at an automatic tool changer. The key constraint is cycle time balance—the robot needs enough time to service all machines without creating idle time on any one.

What about chip and coolant management in the cell?

This is one of the top three issues on every machine tending project. Chips accumulate on fixtures, in vises, and on the part itself. We address it at multiple levels: air blast nozzles in the robot path (2–3 seconds of directed air), programmable coolant wash stations for internal features, and chip confirmation sensors on fixtures that verify the seating surface is clean before the robot loads the next part. Inside the CNC, we spec high-pressure coolant (70+ bar) to break chips effectively and recommend programmable chip flush cycles between parts.

What safety standards apply to machine tending cells?

Machine tending cells must comply with ANSI/RIA 15.06 (industrial robot safety), ISO 10218-2 (robot system integration), and OSHA machine guarding requirements. We design every cell to Category 3 / Performance Level d (PLd) per ISO 13849 for safety-related control functions. Safeguarding options include hard fencing with interlocked access doors, light curtains for part load/unload openings, and area scanners for collaborative robot applications. Every cell ships with a risk assessment document per ISO 12100 and a safety system validation report.

Do you provide training for our operators and maintenance team?

Every machine tending cell includes operator and maintenance training as part of the project scope. Operators learn cell startup/shutdown procedures, material staging, HMI operation, and basic fault recovery. Maintenance personnel receive training on robot jogging, backup/restore, gripper maintenance, and sensor replacement. We also offer advanced robot programming training for shops that want to teach new part numbers in-house. Our maintenance and support team provides ongoing technical support after installation.

What's the typical installation timeline?

From purchase order to production, a standard single-machine tending cell takes 16–20 weeks. Multi-machine cells and rail systems typically take 20–28 weeks. The timeline breaks down roughly as: detailed design (4–6 weeks), fabrication and component procurement (6–10 weeks), assembly and testing at our facility (3–4 weeks), installation and commissioning at your site (1–2 weeks), and production ramp-up support (1–2 weeks). We perform a full Factory Acceptance Test (FAT) at our facility before shipping—you're welcome to witness the test and run your actual parts through the system before it leaves our building. Learn more about our engineering approach on our consulting services page.