How to Verify CNC Machine Repeatability

Why Repeatability Matters for CNC Production

A CNC machine can produce a single good part and still fail in production. Repeatability is the ability to return to the same position and produce the same result cycle after cycle. For engineers and buyers, this is not a theoretical concern. It determines whether a fixture, toolpath, or process will hold up over hundreds of parts.

Many production failures trace back to drift. A spindle may wear slightly. A linear guide may lose pre-load. A thermal expansion issue in a long gantry may shift a part by a few microns after an hour of running. Each of these issues shows up in repeatability tests before it shows up in a customer complaint.

This guide walks through a practical verification sequence. It assumes a working machine and a basic understanding of coordinate systems. The goal is to confirm that the machine can repeat a process with high accuracy and precision, and to identify where the limits lie.

What You Need Before You Start

Before touching the machine, gather the right tools and set the right conditions. Skipping this step leads to data that looks clean but is actually garbage.

Use a calibrated micrometer, a bore gauge, or a CMM if one is available on site. For a quick check, a set of precision gauge blocks and a digital height gauge work well. The tool must have a documented calibration trace. A micrometer that was last calibrated two years ago will not give you reliable deviation data.

Prepare a test workpiece. A simple aluminum or steel block with a flat face and a few drilled holes is ideal. Avoid using a production part as the test coupon. You want a sacrificial piece that can be machined repeatedly without worrying about tool wear on the part itself.

Control the environment. Let the machine and the workpiece sit in the shop for at least an hour before testing. Temperature changes in the air, the machine enclosure, and the workpiece cause expansion and contraction. A machine that was cold at 7:00 AM and warm at 11:00 AM will not measure the same size.

Step 1: Run a Zero-Offset Check

Start by homing the machine. Run the zero offset check on each axis. Move the spindle to a known position, measure the actual position with an external indicator, and compare it to the programmed position.

This step isolates backlash from positioning error. If the measured position drifts as you reverse direction, you have backlash. That drift will show up in every cut. Do not proceed to full repeatability testing until you understand the backlash pattern.

Step 2: Machining a Standard Test Pattern

Load a simple test program. The pattern should include a square, a circle, and a few holes of different diameters. Keep the toolpath simple. Avoid complex contours that mix multiple tool changes.

Run the program three times in a row without changing anything. Do not stop between cycles. Do not re-home. The goal is to measure what the machine does when it repeats the exact same commands.

Measure the first part. Record every critical dimension. Then measure the second part. Then the third. If you only measure one part, you have a data point. If you measure three, you have a sample.

Step 3: Measure Across Multiple Cycles

This is where most shops skip the real work. They measure one part, pass it, and move on. That tells you nothing about repeatability.

Measure each dimension on parts one, two, and three. For each dimension, calculate the spread. If a hole measures 10.000, 10.002, and 10.001, the spread is 2 microns. If it measures 10.000, 10.005, and 10.003, the spread is 5 microns. The second case may still pass tolerance, but it is less stable.

Pay attention to direction. If the hole size changes when you approach from the positive X direction versus the negative X direction, you have a systematic error. That error will not average out over a production run.

Step 4: Perform a Cyclic Position Test

Next, test the machine at multiple positions, not just one. Pick three points across the travel range of each axis. A common setup is the center, a point near the low end, and a point near the high end.

At each point, run the same cut. A small pocket or a face mill pass works well. Measure the depth and the diameter at each point.

This test reveals positioning errors that are not present at the home position. A ball screw with a worn lead will show more error at the end of travel. A thermal gradient in a long axis will make the top of the machine behave differently from the bottom.

Step 5: Apply Statistical Process Control

Collect the measurements from the cyclic test. Put them into a simple chart. A control chart with the mean and the standard deviation will show you if the process is in control.

If the points fall within a tight band around the mean, the machine is repeating well. If the points drift upward or downward over time, something is changing. A tool wearing faster than expected. A coolant concentration shifting. A fixture loosening.

Do not rely on pass or fail. A process can be within tolerance and still trending toward failure. The chart shows the trend before the tolerance is broken.

Step 6: Verify Tool Compensation

Tool wear changes the effective cutting dimension. A worn end mill will cut a smaller hole. A worn face mill will leave a deeper pass. If you are measuring the part but not accounting for tool wear, your repeatability data is misleading.

Run the same test with a fresh tool. Then run it with a tool that has been used for a known number of cycles. Measure both. The difference is your tool wear offset.

This matters because in production, you cannot always start with a fresh tool. You need to know how much compensation to apply and whether that compensation holds over time.

Step 7: Document and Review

Write down what you measured, what you found, and what you did about it. Keep the raw data. Keep the chart. Keep the program version number.

This documentation serves two purposes. First, it gives the next engineer or buyer a baseline. Second, it supports quality audits. If a customer asks how you verified repeatability, you can point to the data instead of a verbal assurance.

Common Mistakes That Ruin Repeatability Data

The most common mistake is testing on a production part. The part has burrs, it has heat, and it may have been clamped unevenly. Use a dedicated test coupon.

Another mistake is changing variables between cycles. If you swap a tool, re-home, or adjust a clamp between part one and part two, you are no longer testing repeatability. You are testing variation.

A third mistake is ignoring the environment. If the shop temperature rises during the test, the metal expands. The part grows. The measurement changes. That is not a machine error. It is a thermal error.

Final Verification Step

After completing the steps above, run one final cycle. Use a fresh tool. Measure the part. Compare it to the first part you measured in the sequence.

If the dimensions match within your expected tolerance and the control chart looks stable, the machine is repeating the process. If they do not match, go back to the cyclic position test. Identify which axis is drifting and which direction the error is moving.

Do not call the machine good until the final part matches the first. That is the whole point. Repeatability is not a one-off result. It is consistency over time.

How This Fits With Other Quality Checks

Repeatability testing works alongside other quality methods. It is not a substitute for CMM inspection of final parts. It is not a replacement for first-article inspection. But it sits underneath them.

If a machine is not repeating, no amount of inspection on the final part will save the process. The inspection will just catch the defect after it happens. Repeatability testing prevents the defect from happening in the first place.

For suppliers, this data is a key part of process control verification. It shows that the machine can hold a process, not just produce one good sample. For buyers, it is the difference between a quote that sounds good and a machine that actually works in the shop.

When to Re-Test

Test after any major change. A new spindle. A new control board. A repair to a way. A change in tooling. A move to a new location.

Test periodically. A monthly or quarterly check is standard practice in most shops. The machine will drift. The data will tell you when.

Quick Reference Table

The table below summarizes the key checks and what they reveal.

Check	What It Reveals	Typical Tool
Zero-offset check	Backlash and basic positioning error	Digital indicator, calibrated micrometer
Standard test pattern	Overall dimensional stability across cycles	Calipers, micrometer, bore gauge
Cyclic position test	Axis-specific drift across travel range	Height gauge, dial indicator
Tool compensation test	Tool wear impact on final dimensions	Fresh and worn tool, same program
Statistical control chart	Process stability over time	Spreadsheet or SPC software
Final verification cycle	End-to-end consistency	Same tools and methods as first cycle