Proof
We test our checker by breaking programs on purpose.
Before a program reaches a machine, our software checks it against the part and the stock. To find out whether that check works, we plant mistakes in good programs and count how many it catches.
Software v0
34 of 51Faults we planted on purpose that our checker caught, across our testsSoftware v0
10Real faults found in 43 programs from outside shopsSoftware v0
14False alarms in those 43 programsAll in simulation. These are the figures on our capabilities page.
Worked example
Our own test plate.
On the drilling program for our own test plate, a vacuum port cover, we planted 10 different mistakes one at a time. The checker caught all 10, and the clean program passed.
- 01A hole moved 5 mm.
- 02A drill that doesn't exist in the tool list.
- 03A rapid move driven into metal that hasn't been cut yet.
Three of the ten. The program is the one behind the drawing at the top of our home page; the check ran on 30 Oct 2026.
Limits
What it doesn't do.
These limits matter as much as the numbers.
- It still misses some faults, such as a reversed spindle, so it doesn't replace prove-out on the machine.
- These are simulation results. No part was cut for this test.
- The checker covers 3-axis programs, and 5-axis programs for the Haas UMC-400, in simulation. It doesn't read tilted work planes or Siemens and Heidenhain 5-axis yet.
- Software checks alone don't establish machining or inspection capability.
Send us a program to check?
Email Jonathan@ariadynamics.us. Tell us about it first and we'll agree how it travels. Please don't email a program for a restricted part.