Sources of error and uncertainty in machine tool calibration Dr Andrew Longstaff Dr Simon Fletcher
University of Huddersfield •
Founded in 1825 • A history of education, innovation and industrial collaboration
• • • •
Over 2,800 staff 24,000 students studying more than 400 degrees An international University • Students from over 130 countries • Delivering courses in China, Hong Kong, India and Singapore
2
EPSRC Centre • The EPSRC Centre for Innovative Manufacturing in Advanced Metrology • based at the University of Huddersfield’s • Centre for Precision Technologies, • a long-established group with an international reputation in precision engineering and metrology research and development.
• Considering concept of the “factory on the machine”, we require elevation of machine tool accuracies • hardware and software solution for stable metrology frame • Efficient, traceable calibration
• High accuracy laser trackers • Sequential multilateration (TracCAL) • Multi DoF lasers • Ultra stable artefacts with scanning/optical probes
Faster
Greater accuracy
Future
Current SoA
Established
Machine tool error measurement
• Synchronised dynamic capture • Affordable simultaneous multilateration • Direct position feedback • Continuous traceable adaptation
Volumetric
Subject of research work 4
Machine tool measurement
5
Main sources of uncertainty • Some significant sources of uncertainty • Measurement methods • Comprehensiveness of data • Measurement of all error sources • Spatial resolution
• • • •
Finite stiffness Thermal deformation Errors due to motion Numerical compensation 6
Measurement methods
International Standards • ISO 230 Test code for machine tools • Part 1: Geometric accuracy of machines operating under no-load or quasi- static conditions • Part 2: Determination of accuracy and repeatability of positioning numerically controlled axes • Part 3: Determination of thermal effects • Part 7: Geometric accuracy of axes of rotation • Part 10: Determination of the measuring performance of a machine tool
• ISO 10791 Test conditions for machining centres 8
Fastest methods can increase uncertainty • For example measurement of nonorthogonality between axes • Using a “squareness artefact” • Should be measured using two straightness measurements and least-squares fit • Often measured with four points on a square.
9
Equipment effects • Environmental factors and operator handling can induce thermal effects in instrumentation. • In particular, angular and straightness interferometry is susceptible to • thermal and • mechanical stresses in the optics • due to the sensitivity of the measurement to beam separation.
• Production constraints often mean that equipment is not given sufficient time to stabilise • Without repeat measurement runs the drift would not be detected.
10
Comprehensiveness of data
Geometric error sources • On a three-axis Cartesian machine there are the traditional “21” sources of geometric error • Some sources are simpler to measure than others • Some have such high measurement uncertainty that they are ignored
B0
Y+ B180 Z+
• On machines with parallel or rotary axes, the number of error sources increases • Neglecting to measure all sources introduces uncontrolled uncertainty 12
Volumetric assessment
Decreasing omission of errors? or Increasing estimated uncertainty?
13
Spatial resolution 50mm steps Error (m)
50 0 -50
0
100
200
300
400
500 600 6.25mm steps
700
800
900
1000
0
100
200
300
400
500 600 4mm steps
700
800
900
1000
0
100
200
300
400 500 600 Axis position (mm)
700
800
900
1000
Error (m)
50 0 -50 50 Error (m)
• Standard resolution measurement of vertical straightness • shows some cyclic behaviour but • may be mistaken as “noise” • not obvious as a ballscrew issue • Sparse data therefore increases uncertainty • However length of test is increased by higher resolution
0 -50
14
Spatial resolution 6
4
Error (m)
2
0
-2
-4
ISO targets (10) Improved targets (20) Fine targets (50)
-6
-8
0
50
100
150
200 250 300 Axis position (mm)
350
400
450
500
3 2 1 Error (m)
• Example 0.5m machine axis • ISO targets efficient • For axes up to 2000 mm, a minimum of five targets per 1000 mm and not less than 5 targets is required by ISO • Error compensation applied on the machine using 10,20 and 50 targets • Residual errors can be significant • Surface finish and form • In reality the minimum is typically used • An effective option is filtering
0 -1 -2
ISO targets (10) Improved targets (20) Smoothed data (20)
-3 -4 0
50
100
150
200 250 300 Axis position (mm)
350
400
450
15
500
Reducing error through filtering • The carriages/feet of an axis act as filters for high frequency imperfections • Mathematical filters can • better approximate this effect • Improve noisy data from, for example, long range straightness data affected by air turbulence • Technique also successfully applied to straightness error compensation • However • Over-filtering may omit real imperfections 16
Finite stiffness
Uncertainty from finite stiffness Fork head
• Long machine table • Significant bend 200µm/m • Uncertainty in
+ve
X axis
A
B Table
Table bend
Error (mic/m)
• reading from a laser interferometry • Squareness of C axis to X
EBX (X pitch error)
250 200
Level A Level B Laser
150 100 50 0
-3500
-3000
-2500
-2000
-1500
-1000
-500
0 -50
Axis position
-100 18
Variation in mass on table •
The machine is often tested “unloaded” • Fixtures and the workpiece can have significant weight
•
The distribution of the mass can cause distortion • Especially if table loads are mounted asymmetrically
•
Own weight, fixture and workpiece mass variation affects geometry • Moving table example showed 90µrad increase in angular error with 1000Kg mass added.
Thermal deformation
21
Machine tool generic heat distributions
Column External heat entering the machine structure
Motor housing (internal heat source)
Carrier
Belt drive
Deformations (internal heat source)of the machine Bearings results in tool displacement relative to the work piece (internal heat source)
Tool (Hence the thermal error) Table Ball screw Base Original machine Deformed machineprofile profile
• Traceability limited to conditions under which calibration made. • Large 5-axis gantry in normal machine shop
70
40 20 0 -20 0
1000
2000
3000 Axis position (mm)
4000
5000
22
6000
Rapid change in temperature 150
• For example probing
• •
• •
Geometric measurements usually taken under nominally stable conditions Rapid error in spindle axial direction from
100 50 Movement (m)
Rapid heating when cutting Rapid cooling when not cutting
0 -50 -100
Y X Z spindle Z boss
-150 -200 -250 0
• spindle speed change or • spindle stop
60
Mechanical and thermal effect Significant error between spindle stop command and probing cycle start
40
200
400
600 800 Time (mins)
1000
1200
50
Movement (m)
• •
1400
16µm/min
M5 Spindle stop command
30 M19 20 Spindle stopped 10
0 0
20
40
60
80
100 120 Time (secs)
140
160
180
200
23
Unforseen thermal effect ylinr epeatm.rtl 10
-10
Error (m)
-20 -30 -40 -50 -60 -70 -80 -90 -2500
-2000
-1500 -1000 Axis position (mm)
-500
0
6 Laser reading (microns)
• Standard axis positioning error measurement • Poor repeatability compared to expectation • Position monitored (EVE) • Cyclic error of >5µm • Due to underdamped air chiller on the machine • Frequency dependent on the environmental temperature • Time to complete measurement approximately 22 minutes • Included just over 1 cycle
Repeatability > 9µm
0
4 2 0
-2 -4
10
30
50
Time (minutes) 24
Errors due to motion Dynamic versus quasi-static
Dynamic data capture •
Dynamic data capture • Not convenient, since actual and nominal positions must be synchronised. • Rarely performed
Bespoke system required
Machine
Z AXIS COMPENSATED ENCODER SIGNAL X AXIS COMPENSATED ENCODER SIGNAL
Z X
CNC Controller
Rotary Encoder
Z AXIS UNCOMPENSATED ENCODER SIGNAL
Encoder interface
X AXIS UNCOMPENSATED ENCODER SIGNAL
Laser interface
•
Measurement Computer
Encoder Interface Compensation Cards
Z
Cross-rail
X
Geometric Compensation Computer
Laser
Linear Enc.
X Axis
Z
26
Dynamic errors - measurement •
•
Cyclic errors and acceleration effects can be detected Dynamic effects can be different with varying • • •
•
location on an axis speeds size of motion
Therefore difficult to quantify
27
Numerical compensation
Numerical compensation •
• • • •
Correction
Error compensation is the process of cancelling or correcting the effect of an error, usually by moving the axes of the machine.
Can correct for errors that cannot X easily be removed by error avoidance. (a) 0
Z X Y
X0 (b)
Error
X0 (c)
Relatively inexpensive and easy to implement when compared to mechanical realignment. Flexible, since it can be updated to accommodate changes in error. Is not intended as a substitute for good mechanical design and build. 29
Numerical compensation methods • Internal (Controller) • Part program modification • Programmable Logic Controller (PLC) • Inbuilt pitch and cross-axis • OEM application • External • PLC • Feedback modification • PC based • Bespoke electronics
• Model types – commercially available • Error map • Simple linear expansion • Parametric equations • HTMs • Model types – ongoing research • Black box • Artificial intelligence 30
Numerical compensation •
Electronic compensation is a powerful tool • Allows repeatable errors to be reduced without expensive mechanical action
But it introduces uncertainty, which varies depending • • • •
upon the skill of the engineer performing the update the quality of the data The quality of the compensation system (An ISO standard is in preparation to help address this issue). 5 0 ANFIS Output Thermal Error Residual value
-5 Thermal drift (micron)
•
-10 -15 -20 -25 -30 -35 -40
31 0
20
40
60
80
100 120 Time (Minutes)
140
160
180
200
Simulation with measured data as a solution
Simulation of effects of errors • Combining all machine error sources allows prediction of errors on probing (and machining) results • Magnitude and direction of errors represented by arrows.
33
The future •
Measurement and simulation • dramatically reducing machine down-time for testing • simulate any conditions for more robust training • seasonal environmental variations, • new running conditions for new products, • agile manufacturing. Range of conditions
Present
Machine downtime
Range of conditions
Machine downtime
Ambition
34
Sources of error and uncertainty in machine tool calibration Thank you
