AI and Industrial Robotics: Machine tending and factory automation

Machine Tending & Factory Automation – Study Guide

What This Is

Machine tending refers to robots or automated systems loading/unloading parts into production machines (e.g., CNC mills, injection molders, presses). It matters because it eliminates repetitive manual labor, boosts throughput, and reduces errors in high-mix, low-volume manufacturing. Example: A car parts supplier uses a collaborative robot (cobot) to load raw aluminum blanks into a CNC lathe, running 24/7 with zero shift changes.

Key Facts & Principles

• Machine Tending (MT): Automated handling of workpieces into/out of production equipment. Example: A robot picks a metal bar from a conveyor, places it into a lathe chuck, then removes the finished part.
• Cycle Time: Total time for one complete tending operation (load-machine-unload). Rule of thumb: MT must match or beat the machine’s processing time to avoid bottlenecks.
• End-of-Arm Tooling (EOAT): The robot’s "hand" (grippers, suction cups, magnets). Example: A 3-finger gripper for irregular parts vs. a vacuum cup for flat sheets.
• Part Presentation: How parts are delivered to the robot (trays, conveyors, bins). Trap: Poor presentation (e.g., tangled parts) causes 80% of MT failures.
• Collaborative Robots (Cobots): Robots designed to work alongside humans without safety cages. Example: A UR10e cobot tending a small CNC mill in a shared workspace.
• Force/Torque Sensing: Sensors that let robots "feel" resistance (e.g., aligning parts). Example: A robot gently presses a gear into a shaft until torque spikes, signaling a fit.
• Changeover Time: Time to switch between part types. Goal: <5 minutes for high-mix production (e.g., using quick-change grippers).
• Safety Standards: ISO 10218 (industrial robots) and ISO/TS 15066 (cobots). Key rule: Cobots must stop if force exceeds 150N (?34 lbs).
• Vision Systems: Cameras + AI to locate/inspect parts. Example: A robot uses a 2D camera to find randomly placed bolts in a bin.
• OEE (Overall Equipment Effectiveness): Metric for automation efficiency: OEE = Availability × Performance × Quality. Target: >85% for automated cells.

Step-by-Step Application

1. Assess the Process
2. Map the current manual tending workflow (e.g., operator loads parts every 30 seconds).

3. Identify pain points: ergonomics, cycle time, scrap rates, or labor shortages.

4. Select the Right Robot

5. Industrial robot: High speed/payload (e.g., FANUC M-10iA for 10kg parts).
6. Cobot: Lower payload (e.g., UR5e for <5kg parts) but easier to redeploy.

7. Rule: Payload = part weight + gripper weight + safety margin (20%).

8. Design Part Presentation

9. Low-mix: Use trays or conveyors (simple, reliable).
10. High-mix: Use bin picking with 3D vision (flexible but complex).

11. Pro tip: Start with trays, then add vision if needed.

12. Integrate with Machines

13. I/O Signals: Robot sends "part loaded" signal to CNC; CNC sends "cycle complete" signal to robot.
14. Safety: Light curtains or area scanners to stop robots if humans enter.

15. Example: A PLC coordinates the robot, CNC, and conveyor.

16. Program & Test

17. Use offline simulation (e.g., RoboDK) to validate paths.
18. Teach waypoints manually or via CAD-to-path software.

19. Test: Run 100+ cycles with worst-case parts (e.g., oily, misaligned).

20. Monitor & Optimize

21. Track OEE and downtime causes (e.g., gripper failures, vision errors).
22. Quick win: Add a "retry" routine for failed picks (reduces stops by 30%).

Common Mistakes

• Mistake: Ignoring part variability. Correction: Test with the worst parts (e.g., warped, oily, or flashy castings). Use force control or vision to handle variation.

• Mistake: Overcomplicating the gripper. Correction: Start with a simple 2-finger gripper. Add complexity (e.g., suction + fingers) only if needed.

• Mistake: Skipping safety risk assessments. Correction: Follow ISO 12100. Example: A cobot tending a press must have a safety-rated stop if a hand enters the danger zone.

• Mistake: Assuming "plug-and-play" integration. Correction: Budget 20–30% of project time for debugging I/O signals and machine interfaces.

• Mistake: Not training operators. Correction: Train staff on basic troubleshooting (e.g., resetting the robot, clearing jams). Pro tip: Create a 1-page cheat sheet.

Practical Tips

• Start small: Automate one machine first, then scale. Example: A shop added cobots to 3 CNC mills before expanding to 10.
• Use modular tooling: Quick-change grippers (e.g., Schunk VERO-S) cut changeover time from 30 minutes to 2.
• Leverage existing infrastructure: Retrofit old machines with I/O modules (e.g., Siemens LOGO!) instead of buying new ones.
• Watch for "automation creep": Don’t add features (e.g., vision, force control) unless they solve a specific problem.

Quick Practice Scenario

Scenario: A medical device manufacturer wants to automate tending for a laser welder. Parts are small (50g), delicate, and arrive in trays. The welder cycle time is 12 seconds. Question: What’s the first step to validate feasibility? Answer: Measure the current manual load/unload time. If it’s >12 seconds, the robot can’t keep up without redesigning part presentation. Why: Cycle time must match or exceed the machine’s processing time.

Last-Minute Cram Sheet

1. Machine tending = robot loads/unloads parts into production machines.
2. Cycle time rule: Robot must finish tending before the machine finishes processing.
3. EOAT (gripper) choice: 2-finger (versatile), suction (flat parts), magnets (ferrous metals).
4. Cobots vs. industrial robots: Cobots = <15kg payload, no cages; industrial = >15kg, cages required.
5. Part presentation trap: Tangled parts = 80% of failures. Use trays or vision.
6. Safety: Cobots stop at 150N force; industrial robots need light curtains.
7. OEE target: >85% for automated cells.
8. Changeover goal: <5 minutes for high-mix production.
9. Trap: Overcomplicating grippers—start simple.
10. Trap: Ignoring I/O signals—budget time for debugging.