What Is Dry Run in CNC Machining?

• 1
  What a dry run in CNC machining actually proves
• 2
  Why the check exists and where it came from
• 3
  When a dry run in CNC machining is not enough
• 4
  How the practice changes on complex machines
• 5
  How to run a dry run in CNC machining on a real machine
• 6
  What each verification method catches

A dry run is a full execution of the program with the cutting removed from the loop. On a machining center we usually raise the Z work offset by 50–100 mm, or set a positive wear offset on every tool, so the toolpath runs at the commanded feed while the tip stays above the part. The control, the post-processor output, the tool changes and the subprograms all run exactly as they will in the real cycle.

What it proves is narrow and valuable. The program parses without alarms. Tool numbers in the T and H registers match the tools that are actually loaded. The turret or the ATC does not collide with the fixture at the safe Z height. Rapids between features land where the CAM system expected them to land.

What it does not prove is equally important. A dry run says nothing about whether the stock material behaves, whether the fixture holds under side load, or whether the tool is long enough to reach the bottom of a deep pocket. It is a geometry and logic check, not a process capability check.

The reason shops with tight tolerance work treat it as standard is cost asymmetry. A program error found at the control costs a few minutes of spindle time. The same error found by the cutter costs a fixture, a workpiece and possibly a spindle.

The practice came from the tape and punched-card era, when a single misplaced block was physically expensive to correct and impossible to preview. Operators ran the tape with the machine in a locked, non-cutting mode to watch the axis motions in sequence. The name stuck even after CAM software gave us backplot and full machine simulation.

Modern simulation is good. It catches interference, over-travel and holder collisions. It also runs on a model, and the model does not know that the vise jaw was swapped, that a soft jaw was skimmed 0.3 mm last week, or that a tool was re-ground to a shorter gauge length. Those are real-machine facts.

So the dry run survives as the last verification layer between the model and the metal. It is not a substitute for simulation; it is what you do after simulation passes and before you trust the setup.

On a 5-axis machine the case is stronger. Rotary and tilt moves bring the holder, the table and the trunnion into each other's space in ways that are hard to see on screen. A dry run with feeds overridden down lets the operator watch the rotary approach with a hand on the feed hold.

• 1
  Simulation checks the modelInterference, over-travel, holder collision, cycle time estimate.
• 2
  Dry run checks the machineActual offsets, actual tool lengths, actual fixture position.
• 3
  Neither checks the materialChatter, deflection and work-hardening only appear under load.

Thin-wall parts are the clearest failure case. A wall 0.8 mm thick in 6061-T6 will move under cutting force even when the toolpath is geometrically perfect. A dry run has no cutting force, so the deflection never appears. For these parts we rely on first-article measurement plus a strategy change, such as reducing radial engagement to 5–8% of the cutter diameter.

Deep pockets and small-diameter tools have the same problem in a different form. A 3 mm end mill at 4× diameter reach will chatter before it breaks, and the chatter depends on spindle speed, tool holder and material. None of that is visible in the air.

Materials that work-harden, such as 304 stainless and some titanium grades, punish a program that rubs instead of cuts. A dry run cannot show you a rubbing condition. You see it as rapid flank wear and a poor finish on the first real part.

Very large parts are a practical limit too. On our 4,000 × 400 × 150 mm travel machines, a full-speed air cut of a long program can take 30 minutes or more. At that point we simulate, dry-run only the critical tool engagements, and probe the datum instead of running the whole file.

On simultaneous 5-axis work, the rotary table and the tilt axis add motion that backplot does not represent well. We dry-run with the tool pulled clear and the rotary moves at 20–30% override so the operator can watch the approach angle into each undercut. The critical check is the clearance between the holder and the trunnion, not the tip.

On a mill-turn center the risk shifts to the transfer between the main and sub-spindle and to bar-puller clearance. A dry run here is usually done with the bar stock removed so the puller can travel its full stroke. Operators watch the synchronization points, where an M-code mismatch produces an alarm rather than a crash.

On a lathe with a tailstock or a steady rest, the dry run must include the approach and retract of those units. A steady rest that closes 2 mm too early will mark the shaft, and the program will never show it.

The common thread is that the dry run becomes a check of the machine's own geometry, not just the program. That is why it stays manual. A skilled operator watching a rotary table index is doing sensor fusion that no simulator reproduces.