The Robot That Dropped the Part Every Third Cycle
A 6-axis robot picked parts from a tray and placed them into a fixture. The robot was properly sized, the reach was fine, the program was clean. Every third cycle, the gripper would release the part 50 mm above the fixture instead of at it. We checked the vacuum generator, the sensor, the program — nothing obvious. Eventually we traced it to the EOAT: the gripper’s vacuum cup was worn, and every third pick was weak. The robot didn’t know it was weak because the vacuum switch was set too loose. It thought it had a good grip until it didn’t.
Robot end of arm tooling (EOAT) is where robot integration succeeds or fails. The robot itself is a commodity — Fanuc, ABB, KUKA, Universal Robots all make good arms. The EOAT is the custom part that touches the product, and it’s where most integration problems live. This article is what I’ve learned about designing EOAT and integrating robots without dropping parts on the floor.
Robot Selection: Payload and Reach Are Not Enough
The customer asks for a robot. The integrator spec’s one. The sales rep recommends a model. The numbers look fine. But robot selection isn’t just payload × reach.
Payload: Don’t Forget the Tooling
The robot’s payload rating includes the wrist load, not just the part. If the part weighs 2 kg and the gripper weighs 3 kg, the robot needs to handle 5 kg plus any dynamic load from acceleration. A robot rated for 10 kg is fine. A robot rated for 3 kg isn’t — even though the part is only 2 kg.
Add 20–30% margin on payload. Robots running at 90% of their rated payload have shorter service life and worse repeatability. You’re not buying a motor; you’re buying a 5-year production asset.
Reach: Account for the Tool Length
The reach number on the datasheet is from the robot’s base center to the wrist flange. Your EOAT extends 200 mm beyond the flange. The part is picked from a tray that’s 100 mm behind the fence. If the robot’s reach is exactly the distance to the flange at the tray position, the EOAT pushes it beyond the rated envelope. The robot either can’t reach or hits a singularity.
Spec the reach with the EOAT length included. If you need 1,200 mm reach to the part, and your EOAT is 200 mm long, you need a robot with 1,400 mm reach (plus margin).
Repeatability: What Do You Actually Need?
A robot that places parts into a fixture needs ±0.02 mm repeatability. A robot that drops parts into a tote doesn’t — ±1 mm is fine. Don’t spec a 6-axis robot with 0.02 mm repeatability for a pick-and-place into a bin. It’s overkill and costs 30% more.
| Application | Typical Repeatability Needed | Robot Class |
|---|---|---|
| Assembly into fixture | ±0.02–0.05 mm | 6-axis, high precision |
| Machine tending (load/unload) | ±0.1–0.5 mm | Standard 6-axis |
| Pick and place to conveyor | ±0.5–1 mm | Standard or SCARA |
| Palletizing into tote | ±1–2 mm | Large reach, low precision |
EOAT Design: The Business End
The EOAT is the part the customer sees every time the robot moves. It’s also the part that touches the product. Design it for the part, not for the robot.
Gripper Type: Match the Part, Not the Catalog
| Gripper Type | Best For | Not For |
|---|---|---|
| Pneumatic parallel gripper | Rigid parts, consistent geometry, high force | Fragile parts, soft parts, variable geometry |
| Vacuum cup | Flat, smooth surfaces (boxes, panels, glass) | Porous, rough, or oily surfaces |
| Magnetic gripper | Ferrous metal parts, flat surfaces | Non-ferrous, thin parts (sag) |
| Soft gripper (pneumatic fingers) | Delicate, variable, food products | High force, heavy parts |
| Two-finger servo gripper | Multiple part sizes, adjustable force | Single part (overkill, cost) |
Sensing: The Robot Needs to Know It Has the Part
Every EOAT needs to confirm it actually picked the part. Without confirmation, the robot runs blind — it might pick air, drop a part mid-motion, or place a missing part. The sensor choice depends on the gripper:
- Vacuum gripper: A vacuum switch that confirms suction before the robot moves. Set the switch threshold so it detects a weak suction (wear on the cup) before the part drops.
- Mechanical gripper: A proximity switch on the gripper fingers that confirms the part is present. Or a force-sensing gripper that detects part presence by the closing force.
- Magnetic gripper: A current sensor on the magnet or a proximity switch that detects the ferrous part.
The sensor signal goes back to the robot controller (or the PLC). If the part isn’t detected, the robot doesn’t proceed — it retries the pick, then alarms. This is the difference between a robot that drops parts on the floor and one that retries and recovers.
The rule: If the EOAT doesn’t have a part-present sensor, it’s not production-ready. The sensor that saves you from a dropped part costs $50. The downtime from a dropped part costs $5,000.
EOAT Construction: Light, Rigid, and Serviceable
The EOAT adds inertia to the robot wrist. A heavy EOAT limits the robot’s speed and cycle time. Make it from aluminum extrusion or carbon fiber, not solid steel. Every gram you save is faster acceleration and a smaller robot.
But it also needs to be rigid. A flexible EOAT that flexes during the move loses repeatability. Use structural aluminum, not sheet metal. Gusset the brackets. If the gripper moves 1 mm when the robot accelerates, your placement accuracy is shot.
Robot Cell Safety: The Fence, the Light Curtain, and the Interlocks
A robot moving at 2 m/s is a hazard. The safety design is not optional.
Guarding Options
- Fixed fence: Aluminum extrusion guarding with interlocked doors. Standard for high-speed cells. The fence prevents access during operation. Opening the door stops the robot.
- Light curtain: A light beam across the work envelope. Breaking the beam stops the robot. Good for load/unload stations where the operator reaches in occasionally. The robot stops safely when the operator enters, then resumes when the beam is clear.
- Safety scanner (area scanner): A laser scanner that defines a safe zone. The robot slows down when a person approaches and stops when they enter the safety zone. Good for collaborative applications without a fence.
- Collaborative robot (cobot): The robot itself has force/torque sensing. It stops when it contacts a person. No fence needed, but slower and lower payload. Good for low-speed, shared-workspace tasks.
Which Safety Level Do You Need?
The safety category (per ISO 13849) depends on the risk assessment. A robot running at full speed inside a fence needs Category 3 or 4 safety — dual-channel safety circuit, monitored e-stop, door interlock. A cobot that moves slowly and has contact detection needs less. Don’t over-engineer the safety (it costs money), but don’t under-engineer it (it costs injuries).
Programming: The Interface That Determines Usability
The robot program isn’t just for the integrator. The customer’s operator needs to change parts, adjust positions, and recover from faults. If the program requires teaching a robot engineer every time, the cell is a maintenance burden.
What the HMI Needs to Do
- Recipe selection: The operator picks the product from a dropdown. The robot loads the corresponding positions, speeds, and gripper settings. No teaching pendant needed.
- Fault recovery: When the robot faults (part dropped, gripper jammed), the HMI tells the operator what happened and how to clear it. “Press reset after removing the jammed part” beats “Robot error code 4023.”
- Position adjustment: The operator can tweak a pick position by ±5 mm from the HMI without entering the robot program. This handles small part variation without a programmer.
- Cycle counter and diagnostics: How many cycles since the last maintenance? When was the last gripper replacement? These go on the HMI, not in the robot’s error log.
Integration Checklist
- What is the total EOAT + part weight? (Informs robot payload, with margin)
- What reach is needed including EOAT length? (Not just to the flange)
- What repeatability does the application actually require? (Don’t over-spec)
- What gripper type matches the part geometry and material?
- Is there a part-present sensor on the EOAT?
- What guarding/safety does the risk assessment require?
- Can the operator run the cell from the HMI without a programmer?
- What happens on a dropped part? (Alarm, retry, safe stop?)
- How will the EOAT be serviced? (Gripper fingers replaceable? Sensors accessible?)
- What is the cycle time target? (Does the robot speed allow it?)
The Bottom Line
Custom robot cell design isn’t about the robot. It’s about the tooling that touches the part, the safety that keeps people safe, and the HMI that lets the operator run it. Spec the robot with margin on payload and reach, design the EOAT with a part-present sensor, guard the cell to the right safety category, and build an HMI that doesn’t require a programmer. The robot that runs unattended for a shift is the one where the EOAT, safety, and HMI were designed as carefully as the robot itself.
