Feed axes on CNC machining centres: wear patterns and their signals
Wear on feed axes first shows up as a loss of accuracy, not as a failure. Backlash and friction are the two quantities worth understanding and comparing over time.
A feed axis rarely fails all at once. Usually it loses accuracy first: a milled circle becomes out of round at the quadrant transitions, a bore sits a few micrometres off after a change of direction, marks appear on the surface at low feed rates. Behind these wear patterns on feed axes lie a few mechanisms in the ball screw, linear guide and bearing arrangement. This article describes how they arise, how to recognise them and which signals from the CNC control can in principle point to them.
Status of our system
Feed axis monitoring is in development with us. In the pilot we record axis signals from the CNC control, but we do not yet evaluate them. There are no alerts and no commitments for this today. What we monitor today is the spindle and its bearings.
How a feed axis is built
Most linear axes on machining centres follow the same principle. A servo motor drives the ball screw via a coupling or a belt drive. The screw shaft is held axially in a fixed bearing at one end and is often guided in a floating bearing at the other. The nut converts the rotation into linear motion and pushes the slide. The slide runs on profile rail guides, in whose carriages balls or rollers recirculate.
The position is measured either by the encoder in the motor (indirect measuring system) or by a linear scale directly on the slide (direct measuring system). The position control loop in the CNC control compares the commanded and actual positions and sets the motor current so that the slide follows the path.
Typical wear patterns on feed axes
The damage mechanisms have long been known. Which of them dominates on a particular machine depends on design, load, lubrication and operating conditions.
| Component | Mechanism | Typical effect |
|---|---|---|
| Ball screw, nut | Wear of the raceways, loss of preload | Backlash grows, stiffness falls, form errors after changes of direction |
| Ball screw, raceway | Pitting, scoring, consequential damage after a collision | Uneven running, noise, fluctuating motor current at certain positions |
| Fixed bearing of the screw shaft | Bearing wear, loosened bearing lock nut | Axial play that acts like backlash |
| Linear guide | Lack of lubrication, worn wipers, ingress of chips or coolant | Friction rises, often unevenly along the travel, stick-slip at low feed rates |
| Coupling, belt drive | Loosening, wear, falling belt tension | Play or compliance between motor and screw shaft |
| Screw shaft, thermal | Heating due to friction in nut and bearings | Change in length that reaches the workpiece as a position error with indirect measurement |
A collision deserves special attention. It can damage the ball screw, fixed bearing and guide carriages without the axis attracting attention immediately. The consequences often only show weeks later. It is worth recording a reference value after every crash.
Backlash and friction: the two basic quantities
Almost all wear patterns can be traced back to two quantities.
Backlash (also called lost motion) is the distance the drive travels on a change of direction before the slide moves. It is caused by play in the nut, fixed bearing and coupling, and by elastic deformation. The CNC control can compensate a fixed amount via backlash compensation. However, this only hides the problem as long as the backlash stays constant. If it grows or varies along the travel, the compensation value no longer fits. The result is steps on circles and contours at the points where an axis reverses direction.
Friction is the resistance the drive has to overcome. Part of it is intended, for example through the preload in the nut and guide carriages. If friction rises, the motor needs more current for the same motion. Often more important than the mean value is the uniformity: a spot with increased friction on a guide rail or a damaged raceway in the screw shows up as a local increase that recurs at the same position on every pass. If preload decreases due to wear, friction can even fall while backlash rises. A falling current is therefore not automatically a good sign.
Which CNC control signals can show wear on feed axes
For every axis, the CNC control knows quantities that are in principle suitable for trend monitoring. Which of them can be read via an interface on a particular machine has to be checked for each control and configuration.
Following error
The following error is the difference between the commanded and actual position during motion. For a purely proportional position control without feedforward, the following applies approximately:
For wear monitoring it is precisely this remainder that is interesting. With feedforward switched on, changes in friction and backlash show up mainly as short peaks in the following error on changes of direction and as altered values when starting off. The absolute following error says little. What is meaningful is its trend for an identical motion over weeks and months.
Motor current, load and torque
The current of a servo motor is approximately proportional to its torque. At constant speed without machining, the load consists essentially of friction and, on vertical axes, the weight of the slide. Rising current for the same move points to more friction. A current that regularly shows peaks at certain positions points to a local cause in the guide or screw. Depending on type, CNC controls provide values such as current, load in percent or torque for this.
Direct and indirect measuring system
If an axis has both a motor encoder and a linear scale, the difference between the two positions is a valuable signal. It contains the play in the drive train and the thermal expansion of the screw shaft. If this difference changes over time for the same move, something in the mechanics between motor and slide has changed.
Temperature
Motor temperatures can be read out on many CNC controls. Above all, they help put measurements in context: a cold axis has different friction from a warm one. The control does not usually supply a temperature at the nut or the fixed bearing. That would require a separate sensor.
Limits of CNC control data
Via standard interfaces such as OPC UA, CNC controls supply values at intervals ranging from tenths of a second to whole seconds. That is enough for trends over weeks. Short events such as the following error peak at a reversal of direction are only captured with luck. For such details, controls have their own trace or oscilloscope functions with much higher resolution. How to connect a control read-only is described in the article Connecting SINUMERIK and HEIDENHAIN via OPC UA.
Comparing trends correctly: same motion, same conditions
Axis signals during normal production are hard to compare. Every program travels different paths at different feed rates, the cutting forces overlay the friction, and the temperature changes over the shift. A rising mean current over a week may simply mean that different parts were produced.
The cleanest approach is a fixed reference run: a short NC program that moves each axis in exactly the same way every time, without machining. Not to be confused with homing at switch-on, which only determines the zero point. These rules have proven themselves:
| Rule | Why |
|---|---|
| Same program, same feed rates, fixed override | Only then are current and following error directly comparable. |
| Full travel where possible, both directions | Local damage only shows at its position, backlash only at reversal. |
| Several speeds, including a slow one | Friction and stick-slip are most noticeable at low feed rates. |
| Same thermal state, for example after a fixed warm-up | Lubricant and mechanics behave differently when cold and warm. |
| Store values by position and direction, not just as a mean | A mean value smears out precisely the local anomalies. |
| Note maintenance and interventions with the date | After lubrication or replacement a new reference state begins. |
If you run this reference move regularly and overlay the results, you will see changes that get lost in production operation. What matters is the comparison with the axis's own initial state, not with a general limit value. This is the same principle we follow on the spindle with the learned baseline, i.e. the normal state of that spindle. More on this in the article Fixed thresholds or learned baseline.
What maintenance engineers can check and measure themselves
Many wear patterns on feed axes can be narrowed down with on-board tools and standard measuring equipment:
- Backlash with a dial indicator: approach the axis in one direction, set the dial indicator, move a short distance in the opposite direction and read off when the slide moves. Repeat at several positions, because backlash can vary along the travel.
- Circularity test: a ballbar makes reversal spikes, backlash and mismatch between two axes visible. The method is described in ISO 230-4. Repeated regularly, it gives a good trend.
- Positioning accuracy: a laser interferometer according to ISO 230-2 measures positioning errors and repeatability along the axis. This is usually a job for a service provider, but a valuable reference.
- Lubrication: check the level and function of the central lubrication system, and inspect the lines to the nut and guide carriages for breaks or blockages.
- Wipers and covers: damaged wipers on the guide carriages and leaking telescopic covers let chips and coolant reach the raceways.
- Sound and feel: during slow manual jogging, listen for uneven running, chattering or knocking at certain positions.
- Compensation values: how have backlash compensation and leadscrew pitch error compensation changed since commissioning? A value that has been readjusted several times is a trend in itself.
The dimensions on the workpiece are also a signal, often the first. If quality assurance reports form errors on circles or offsets after changes of direction, it is worth looking at the axis before changing the program.
What we do today and what is still missing
In the pilot we read axis signals such as load and following error, as far as the CNC control provides them. We store them in order to understand how stable they are in real operation and how comparable conditions can be recognised automatically. There is no evaluation and there are no alerts for them yet. We will only offer them once we have shown on real machines that the conclusions are reliable. Access does not change in any way: our system only reads and writes nothing to the control.
Today we monitor the spindle and its bearings with a dedicated vibration sensor and an assessment against the learned baseline. How this works is explained in the guide to spindle monitoring. The current scope is described on the page Spindle monitoring for CNC milling centres.
Frequently asked questions
How do I recognise wear on a ball screw?
Typical signs are growing backlash, form errors on circles at changes of direction, uneven running and noise at certain positions. This can be measured with a dial indicator, a circularity test and a laser interferometer. In the CNC control data it shows up as changed motor current or following error during an identical reference run.
What does a rising following error on a CNC axis mean?
For the same motion and unchanged controller settings, it points to more friction, more backlash or a changed load. On its own it is not a finding. First check whether the program, feed rate, controller parameters and temperature were comparable.
Is the CNC control's backlash compensation sufficient?
It compensates a fixed amount and is intended for small, constant backlash. If the backlash grows through wear or changes along the travel, the value is no longer correct. A compensation value that is frequently readjusted is a sign to check the mechanics.
Can linear guides be monitored during operation?
In principle yes, via the motor current during identical moves and via vibration or noise measurements on the guide carriage. What matters is the comparison with the guide's own initial state. Checking lubrication and wipers nevertheless remains the most important preventive measure.
Does Forbach & Partners monitor feed axes?
Not yet. Axis monitoring is in development. In the pilot we record axis signals but do not yet evaluate them. Today we monitor the spindle and spindle bearings.
About the author
Sascha P. Forbach is the founder of Forbach & Partners and the software architect of its spindle monitoring system. For twelve years he has advised manufacturers on process quality and machine availability. About us