Fixture Repeatability Check After Removal from the Table

Where the problem shows up

The problem usually does not appear after a breakdown, but during normal changeovers. The fixture is removed from the table because a different batch arrived, a rush order came in, or space was needed for another setup. A day or a week later, it is put back, the fasteners are tightened, and everyone expects the same zero as before.

On paper, that sounds logical. On the shop floor, it does not always work that way. The fixture comes back almost to the same place, but not to the exact same point. A little chip was left on the base, someone tightened the bolts in a different order, the support surface picked up a burr, or the key seated slightly differently. Every small detail adds a shift.

There is another trap too. The operator sees that the program is the same, the part offset is still there, and the coordinates look familiar. It seems enough to put the fixture back on the table and keep running. But a saved zero does not correct a new seating position. If the fixture itself sits 0.02-0.05 mm differently, the machine will faithfully cut from the wrong base.

That kind of shift is easy to miss. Two or three hundredths at setup may look harmless, but for a finish dimension, a hole in a second operation, or a bearing fit, that is already enough. The part looks fine, and then inspection finds it out of tolerance.

That is usually where the argument starts. The operator says the fixture was returned by the marks. The setter suspects the machine. Inspection sees the size drift but cannot point to the source right away. Without a simple test, it is impossible to tell whether the cause is table repeatability, the fixture itself, or the way it is positioned each time.

For shops that change fixtures often on CNC lathes and machining centers, this is a normal working situation. And it is solved not by arguing, but by a short check with one base part and several repeat setups.

What to prepare before the check

Before the test, remove every random source of variation. The idea is simple: only one condition should change — you remove the fixture from the table and put it back. If the part, clamping force, or measuring method changes as well, the result can no longer be trusted.

Start with one simple base part made from a stable material. A blank without thin walls and without a shape that easily moves under clamping is best. You need clear support surfaces and several points that are easy to reach with an indicator or feeler gauge. The simpler the geometry, the easier it is to tell where the error comes from: the table, the datum, or the part itself.

Use the actual fixture from production, not a training setup. Otherwise you will be testing a nice drawing, not the real shop-floor situation. On a machining center, something as small as chips in a slot or a burr on a support often causes more drift than expected.

For all three setups, keep the same kit:

• one base part that does not change during the test;
• a fixture in working condition, without disassembly between cycles;
• the same tool, clamping method, and measuring approach;
• an indicator, a gauge, and a simple measurement log;
• a clean table, slots, supports, and contact surfaces.

Prepare the log in advance. A short table is enough: date, who set it up, where it was measured, and what deviation was found. It saves time and quickly shows whether the error repeats or goes in a different direction each time.

Also check the clamping method. One operator tightens harder, another lighter, and then you are no longer comparing table repeatability, but differences in force. If you have a torque wrench, use it. If not, at least keep the same order of operations.

Before the first indicator touch, wipe the table, the slots, the lower face of the fixture, and all support points. Five minutes of cleaning is often more valuable than a long debate about machine accuracy.

How to choose a base part

For this test, you do not need a complex production part. It is better to use a simple blank that does not bend under clamping and does not hide table movement with its own deformation. Complex shapes only get in the way here.

The most convenient option is a compact block made from a familiar material. Choose a part without thin walls, deep pockets, or long overhangs. If the blank flexes or vibrates during machining, you will be catching the part’s behavior, not the remounting error.

A good test part is usually short and rigid. It has 2-3 machined surfaces that are easy to reach with a gauge, indicator, or micrometer. It helps if it includes a dimension that changes even with a small X or Y shift.

In practice, a very simple geometry is often enough: a top face, two side faces, and one shallow slot or pocket referenced from those faces. Then, after each removal and return, you can check not only the overall size but also the position of the feature relative to two sides. If the fixture does not return correctly, that dimension will move immediately.

A good example is a rectangular block 80 x 60 x 25 mm. Machine the top, front, and right sides, then cut a shallow pocket near the corner. After each removal and return of the fixture, measure the distance from the pocket walls to the two side datums. A shift of a few hundredths is much easier to see on a part like this than on a complex contour.

Do not mix this check with a full process validation. The goal here is narrower and simpler: find out quickly whether the fixture holds the same position after being put back on the table. So the base part should machine quickly, measure quickly, and give a clear result.

How to run the test in three setups

You do not need a complicated route. You need one base part, the same fixture, and strict discipline: change only one thing — remove the fixture from the table and put it back.

First clamp the base part and machine 2-3 reference surfaces or dimensions that are easy to measure after each setup. It is better to choose elements where the error shows up right away: a face, a hole, the distance between two surfaces, or runout on the outer diameter. Do not add new operations during the test. If the cutting conditions, tool, or clamping scheme change, the test will no longer tell the truth.

Then keep the same sequence every time:

1. Mount the fixture on the table, load the part, run the cycle, and record the measurement result.
2. Remove the fixture completely from the table. Do not loosen it only halfway and do not shift it just for show.
3. Clean the table, keys, support areas, and the fixture contact surfaces.
4. Put the fixture back on the table, tighten it the same way, load the part again, and repeat the same cycle.
5. Do this one more time so you have three separate setups and three sets of measurements.

During the test, do not change anything else. Keep the same cutting conditions, the same tool, the same overhang, the same program, the same tightening torque, and the same part datum sequence. If in the second or third setup the operator starts clamping the blank “a little more carefully,” then you are already testing human action, not table repeatability.

A good sign is when the spread between the three setups stays small and within a narrow range. If the value moves in a new direction after every removal and return, the cause is usually in the seating surfaces, contamination, key wear, or the datum scheme itself.

What to measure after each setup

After each removal and return of the fixture, do not look at every dimension. Focus on a few fixed points. That makes it easier to see whether the seating on the table is shifting or whether the problem is in the part itself.

Start with the same X dimension every time. It should be a simple dimension from a stable datum to the same surface or axis. Do not change the measurement point between the first, second, and third setup, or you will compare different inspection methods instead of setups.

Then check the same fixed Y dimension. If X holds and Y drifts, the cause is often a side stop, key, contamination at the seating surface, or a tilt during tightening.

If there is a risk of vertical movement, add a Z check. Measure the height from the datum to the same face of the part or a control feature on the fixture. This is useful when the table, adapter plate, or the fixture itself may not seat exactly the same way after remounting.

Also check a reference surface with an indicator. Sweep the indicator across the same surface every time and record not only the maximum deviation, but also the pattern of the reading. If the needle always moves in the same place, look for a local cause. If the pattern changes from setup to setup, the seating is drifting.

Usually four values are enough after each cycle:

• the X dimension;
• the Y dimension;
• the Z height, if it matters;
• runout or deviation of the reference surface from the indicator.

Then compare the first, second, and third setup. Look at the spread, not just the size itself. If the first and second setups match and the third moves away, that is already a signal: either the seating is unstable, or the fixture is being tightened differently.

The most practical method is to put the data into a short table and immediately calculate the difference between the minimum and maximum values. For example, if you get 40.012, 40.014, and 40.013 mm on X, the spread looks calm. But if the tolerance is 0.05 mm and the size is already moving by 0.03 mm, most of your accuracy reserve is gone before machining even starts.

If one dimension drifts while the others stay steady, the problem is local. If X, Y, and the indicator reading all shift, it is usually the fixture location after remounting.

A shop-floor example

A small flange part was being machined in a production batch on a CNC lathe. The part was simple, but the tolerance was tight: the dimension from the reference face to the seating surface had to stay within hundredths.

To avoid arguing about the cause of rejects, the foreman ran a short test. They used one base part, the same fixture, and repeated one scenario three times: mounted it, machined the control dimension, removed the fixture from the table, and put it back in place.

The result was very clear:

• after the first setup, the size was on target;
• after the second return, it moved positive by about 0.01 mm;
• after the third return, the positive drift reached 0.03 mm.

At first, they suspected the program. That is a normal reaction: if the size drifts, someone may have touched the correction, or the error may be in the cycle. But here the logic was different. If the program were the cause, the size would have shifted immediately and repeated the same way on every new part. Instead, the deviation grew only after the fixture was removed and returned.

They checked the simplest things first. The fixture was removed again, the table, support areas, and lower face were wiped clean, and then an indicator check was done. Under one support, they found dirt and a thin chip that was almost impossible to see by eye. That was enough to make the fixture sit slightly higher every time.

After cleaning, the test was repeated with the same setup. The size returned almost to the original point, and the spread between setups dropped to a few microns. The program, tool, and offsets had nothing to do with it.

This kind of check is useful because it quickly separates datum errors from machining errors. When the size moves in steps after each remount, start by looking for dirt, burrs, a tilt during tightening, or support wear. In that situation, there is usually no need to adjust the program.

Where people usually go wrong

The table is often blamed too early. In practice, the issue often appears not in the machine, but in the test itself: the conditions change, the part is awkward, or small details on the supports are ignored. Then the test no longer shows the real picture.

The first common mistake is using a base part that is too complex. If the part has many operations, thin walls, or several questionable datums, you end up testing everything at once: the part, the clamping, the program, and the material behavior. For this test, you need a simple part with clear geometry and a stable datum scheme.

Different clamping force between setups causes just as much confusion. One time the operator tightened the fixture harder, the next time softer, and the third time they also changed the tightening order. Then the deviation gets blamed on the table. That is not correct.

Another typical mistake is comparing different parts. Sometimes after the first setup one blank is measured, after the second a different one is used, and after the third someone checks a part from another batch. That destroys the common reference point. You need one base part and one control method.

People also often forget to record temperature and time after the machine warms up. Then they wonder why the result is one thing in the morning and another an hour later. If the spindle, table, and fixture reached operating temperature at different times, the numbers will drift. It is enough to record at least the warm-up time and the conditions under which each measurement was taken.

And there is one very simple cause of errors: chips, dirt, and supports. One small curl under a support foot can create more shift than the table assembly itself. The same happens with a ding on the base, an oil film, or a misaligned adjustable support.

Before you blame the table, check four things:

• are the datum surfaces and supports clean;
• was the fixture tightened the same way each time;
• did you use the same part and the same measurement method;
• did you record warm-up time and measurement conditions.

This order often saves several hours and quickly shows where the error really comes from.

A quick check before you draw conclusions

Before deciding whether the fixture keeps repeatability after removal, check one simple condition: all three setups must be done the same way. One missed burr or a changed offset can ruin the whole test.

A quick check takes only a couple of minutes:

• clean the table, supports, and lower face of the fixture;
• make sure the fasteners were tightened in the same order in all three setups;
• check that the tool, overhang, and offsets stayed the same;
• compare the blanks if the test uses a batch rather than one part;
• put the three results next to each other instead of leaving them in separate shift notes.

If even one point does not match, do not rush to blame the table. Repeat the cycle under the same conditions first. Otherwise, the check turns into a collection of random deviations.

People are more often confused not by measurement, but by small inconsistencies in the process. On the first setup, the operator tightened the bolts in a cross pattern; on the second, they went around in a circle; on the third, they quickly tightened only two fasteners. Formally, it is the same test. In reality, the conditions are already different.

Recording the results matters too. When the dimensions are scattered across different notebooks or kept in the memory of two operators, the picture breaks apart. It is much easier to look at three numbers side by side: 12.006, 12.008, and 12.007. That spread looks calm. But if the third value becomes 12.041, first look for dirt under the fixture, a tilt during tightening, or a change in offset.

How to read the result

The test only makes sense when the acceptable spread is known in advance. Without that, the numbers look either “fine” or “bad” by eye, and that almost always leads to an argument. A simple rule works well: repeatability should take up only a small part of the machining tolerance.

If the part tolerance is ±0.02 mm and the spread after three setups is already 0.01 mm, the safety margin disappears too fast. Even if the parts still pass inspection, there is not much room left for tool wear, heat, and normal variation during production.

In practice, it helps to look at the result this way:

• the spread is clearly below the internal repeatability limit — the fixture can go into production;
• the spread is close to the limit — the setup is still usable, but it needs more frequent checks;
• the spread is above the limit or grows from setup to setup — the cause must be found before production starts.

When the result is poor, do not rush to blame the machine table itself. First check the datum, clamping, supports, and the condition of the contact surfaces. A single chip under a support or a slightly uneven tightening force can create an error faster than it seems.

It is useful to look not only at the size of the spread, but also at its pattern. If the size moves in one direction every time, the cause is often in the stops, keys, base seating, or a tilt during tightening. If the deviations are random, it is more often contamination, different clamping force, or wear at the support points.

If the test shows repeated instability, include it in the normal fixture acceptance process. This is especially useful for fixtures that are often removed from the table, moved to another machine, or changed between batches. A short log with the date, fixture number, and three results quickly shows whether the problem is systematic or just a one-off.

The working routine is simple: set a permissible spread for each fixture type, record the result of three setups, check the bases, supports, and fasteners immediately if the limit is exceeded, and repeat the same test after the fix. It is better to release the fixture into production only after a stable repeat result.

If you are choosing a new machine or launching a new fixture, it is better to agree on these checks in advance. EAST CNC, the official representative of Taizhou Eastern CNC Technology Co., Ltd. in Kazakhstan, handles not only machine supply, but also commissioning and service, so it is convenient to include such tests in the acceptance process right away. That saves you from searching for the cause later during production.