SISOM 2004, BUCHAREST, 20-21 May

ERROR PREDICTION AND COMPENSATION FOR CNC MACHINE TOOLS

Doina MARIN

Scientific researcher, Institute of Solid Mechanics Romanian Academy, e-mail: [email protected]

This paper presents the development of kinematics error models accounting for geometrical and
thermal errors in the Vertical Machining Center with open architecture controller. Using Rigid Body
Kinematics and small angle approximation of the errors, each slide of the three axes vertical
machining center is modeled using homogenous coordinate transformation. The developed
mathematical model is used to calculate and predict the resultant error vector at the tool-workpiece
interface for error compensation.

1. INTRODUCTION

In order to fulfil the requirements of modern industry, the accuracy of multi-axis machines has been of
the principal concerns in their design and implementation. A number of methods for the analysis, prediction
and compensation of geometric errors have been proposed. All methods have the common objective of
defining the error in the nominal relationship between the tool and the workpiece, and utilize a variety of
methods to achieve this objective. Some of these are presented below.
Based on the assumption that both the machine and workpiece are rigid bodies. Anjanappa, et al. [1]
developed two methodologies for cutting force independent error compensation. In the first, the machining
codes to the machine controller are intercepted and changed to facilitate compensation, and the second, the
tilt and translation of a magnetic bearing spindle is utilized to perform real time correction. Zhang [12]
presents results and methods for error compensation of a three-axis coordinate-measuring machine.
Compensation for systematic geometric errors and first order expansions are incorporated in the technique.
Kiridena and Ferreira [4,5,6] published a series of papers that include an error model, model parameter
measurement, and compensation schemes for quasi-static errors using rigid body kinematics.
Although extensive research has been conducted to derive a volumetric error model most of the works
has focused on turning centers. The identification of volumetric error can be classified into two alternative
methods: the direct workspace identification method and the error synthesis method. The first approach
directly identifies the volumetric error of 3D machine by using artifacts with known dimensions placed on
the machine. The machine using a touch trigger probe then measures the artifacts. The difference between
measurement and the known lengths are used to calculate the volumetric error. Because of the limitations of
the direct identification method and progress in laser interferometer measurement techniques, recent research
has focused on the error synthesis method.
The predicted errors are incorporated into a software-compensation scheme. The underlying idea is
very straightforward and feasible. Since the nominal cutting path is given in a form of NC code, the
compensation software identifies the coordinates of the cutting trajectory and used them as the inputs for
error predictions. The cutting trajectory will then be automatically modified according to predicted position
errors of tool. Finally, the new NC codes will be re-generated and issued to the machines CNC controller for
machining.
101 Error prediction and compensation for cnc machine tools

This paper presents the derivation of a general volumetric error model, which synthesizes both
geometric and thermal errors of a three-axis vertical machining center using homogenous matrix
transformations of the axis slides. The developed mathematical model will be used to calculate and predict
the resultant volumetric error vector at the tool workpiece interface for error compensation.

2. MACHINE TOOL ERRORS

The accuracy of the machine tool is primarily affected by the geometric errors caused by mechanical-
geometric imperfections, misalignments and wear of the linkages and elements of the machine tool structure,
by the non-uniform thermal expansion of the machine structure and static/dynamic load induced errors. As a
result a volumetric error which is the relative error between the cutting tool and the workpiece is created. The
errors can be reduced with the structural improvement of the machine tool through better design and
manufacturing practices. Due to physical limitations, production and design techniques cannot improve
accuracy. So, identification, characterization and compensation of these error sources are necessary to
improve machine tool accuracy cost effectively.
In machine tools there are three main sources of errors:
errors due to geometric inaccuracies;
thermal errors generated by environmental temperature changes, local sources of heat from drive motors,
friction in bearings and other transmission devices, and heat generated by the cutting process;
load induced errors (workpiece weight, forces resulting from cutting process, gravity forces resulting from
the displacement of masses of the machine components).
Geometric errors are regarded as the machine tool errors, which exist under cold start conditions. The
mechanical imperfections of the machine tool structure and the misalignment of machine tool elements cause
them. They all change gradually due to component wear. The effect of the geometric inaccuracies is to
produce errors in the squareness and parallelism between the machines moving elements.
The 21 geometric error components of a three axes vertical machining center are presented in table 1.

Table 1
Errors Number of error components
Linear positioning errors (scale error) 3
Straightness errors 6
Angular errors 9
Orthogonality (squareness) errors of machine axes 3

In figure 1 a schematic of six error components of the vertical machining centers X - axis carriage
system is presented. Notational conventions are used: OXYZ reference coordinate system; O1X1Y1Z1
carriage coordinate system; X desired direction of motion; - angular error motion; - the translation error
motion. A subscripted letter represents the axis the motion rotates about and a letter in brackets represents
the motion direction. Positive rotation is defined by the right-hand rule.

Fig.1 Schematic of six degrees of freedom error motion of a machine tool carriage
system
 Doina MARIN 102

3. A BRIEF OVERVIEW OF HOMOGENEOUS COORDINATE SYSTEM TRANSFORMATION

Using Rigid Body Kinematics, each axis of a machine tool relative to each other and to the reference
frame can be modeled using a Homogenous Transformation Matrix (HTM). A Homogenous Transformation
Matrix in 3 dimensional space is a 4 x 4 matrix. It can be used to represent one coordinate system with
respect to another or reference coordinate system. Generally speaking, this matrix has the form presented
bellow:

n x sx ax Px
n sy ay Py n s a p
1
T2 = y = (1)
nz sz az Pz 0 0 0 1

0 0 0 1

where: n, s, a represent orientation of a coordinate system with respect to another coordinate system and p
represents the position vector of the origin of a coordinate system with respect to another coordinate system.
The pre-superscript represents the coordinate frame we want the results to be represented in, and the post-
subscript represents the coordinate frame we are transferring from.
In case of carriage-guide system we can write the homogenous transformation matrix. The locations of
the workpiece and cutting tool are established relative to the machine tool reference frame and the resultant
volumetric error is estimated. A HTM describes the pure translation of an ideal carriage (fig.1) for the X -
axis:

1 0 0 x + a
0 b
[ T]
R
1 ideal =
0
1 0
0 1 c
(2)

0 0 0 1

where x represents variable position of the carriage coordinate system origin with respect to the reference
coordinate system, a, b, c represents constant offsets of the origin of the carriage coordinate system
(O1X1Y1Z1) origin with respect to the reference coordinate system (OXYZ) in x, y, z directions, respectively.
The total error motion of a carriage is a combination of rotation and translation errors. HTM of rotation and
translation errors are given below with the assumption of small angular errors:

1 z y 0
1 x 0
Trot = z ; (3)
y x 1 0

0 0 0 1

1 0 0 x
0 1 0 y
Ttrans = (4)
0 0 1 z

0 0 0 1
103 Error prediction and compensation for cnc machine tools

where: x (roll), y (pitch), z (yaw) are rotational errors about x, y and z axes respectively. x , y and z
represent translation errors along X, Y, Z axes, respectively (fig.1). The resultant HTM describing the error in
position of the carriage with respect to its ideal position is:

1 z y x
1 x y
E= z (5)
y x 1 z

0 0 0 1
The actual position and orientation of the carriage in reference coordinate system can be calculated as
following:

1 z y x + x + a
y + b
[ T]
R
1 actual [ ]
= T1 ideal E = z
R
y
1
x
x
1 z +c
(6)

0 0 0 1

[ ] [ ]
where R T1 actual represents actual HTM of the carriage in reference coordinate system, R T1 ideal represents
ideal HTM of the carriage in reference coordinate system and E represents error HTM of the carriage.

4. DERIVATION OF ERROR MODELS FOR A THREE AXES MACHINE

In order to specify the nominal position of any point in space, a common reference should be imposed
on so that a concept of volumetric errors can be realized in terms of this reference. The actual position and
orientation of the X axis carriage of a three axes machine (fig.2) in Y axis carriage coordinate system is:

Fig.2 Schematic of a three axis machine tool

1 z 2 (x ) y 2 (x ) x + x 2 (x ) + a 2
(x ) x 2 (x ) y 2 ( x ) + b2
[T ]
1
2 actual = z2
1
y 2 (x ) x 2 (x ) 1 z 2 (x ) + c 2
(7)

0 0 0 1

where x2(x), y2(x), z2(x) are roll error, pitch error and yaw error of X- axis respectively; a2 ,b2 ,c2 are
constant offset in X, Y, Z directions between O1 i O2 ; x2(x) is linear displacement error of X axis.

y 2 (x ) = y' 2 ( x ) + xy x (8)

where: y2(x) Y straitness of X axis as it moves in X direction; xy orthogonality error between X and Y
axes; x nominal X axis position that amplifies xy to yield an Abbe error in Y direcrtion.

z 2 (x ) = z' 2 (x ) + xz x (9)

where: z2(x) Z straightness of X axis as it moves in Y direction; xz - orthogonality error between X and
Z axes; x nominal X axis position that amplifies xz to yield an Abbe error in Z direction.
In the same way, it is defined the actual position and orientation of the Y and Z axis carriage in
reference coordinate system (fig.2):

1 z1 ( y ) y1 ( y ) x1 ( y ) + a1
(y) x1 ( y ) y + y1 ( y ) + b1
[ T]
R
1 actual = z1 1
y1 ( y ) x1 ( y ) 1 z1 ( y ) + c1
(10)

0 0 0 1

1 z 3 (z ) y 3 (z ) x3 (z ) + a3
(z ) x 3 (z ) y 3 (z ) + b3
[T]
R
3 actual = z3
1
y 3 (z ) x3 (z ) 1 z + z1 (z ) + c3
(11)

0 0 0 1

Let 3Ttool be the cutting adge of the cutting tool, which is attached to the spindle. The vector 2Twork
represents the cutting adge at the workpiece which is fixed on the worktable (fig.2). These two cutting edges
can be transformed to the reference frame using HTM. In the ideal case, there are no linkage errors. As a
result, these two cutting edges must be coincident. Thus, we have:

R
Twork actual = R T1 actual 1T2 actual 2Twork (12)

R
Ttool actual = R T3 actual 3 Ttool (13)

R
Twork = R Ttool (14)

2
Twork and 3 Ttool are represented by the following vectors in their respective coordinate frames:

2 T
Twork [
= Wx W y Wz 1 ] (15)
105 Error prediction and compensation for cnc machine tools

3 T
Ttool [
= Tx T y Tz 1 ] (16)

where Wx ,Wy and Wz are workpiece coordinates; and Tx ,Ty and Tz are the X, Y, Z offsets of the tool tip,
respectively. T reprented transpose of the matrix. For the ideal case, transformation matrices of R T1 , 1T2 i
R
T3 are given as follows:

1 0 0 a1 1 0 0 a2 + x 1 0 0 a3
0 1 0 y + b1 0 1 0 b2 0 1 0 b3
R
T1 = 1
T2 = R
T3 = (17)
0 0 1 c1 0 0 1 c2 0 0 1 c3 + z

0 0 0 1 0 0 0 1 0 0 0 1

where a1 ,b1 ,c1 are constant offsets between O and O1; a2 ,b2 ,c2 are constant offsets between O1 and O2; a3 ,
b3 and c3 are constant offsets between O and O3; x,y and z are the nominal axis positions.
After carrying out the matrix multiplication in (12) and (13) and substituting into (14), the workpiece
coordinate vector can be calculated as follows:

2
Twork = [T
R
1
1
T2 ]
1
[TR
3
3
Ttool ] (18)

W x a1 a2 + a3 x + Tx
W b b2 + b3 y + T y
2
Twork = y = 1 (19)
W z c1 c2 + c3 z + Tz

1 ideal 1 ideal

The 21 parametric errors will cause a relative error between tool and workpiece because the linkagees
are not perfect .
The actual spatial relationship between the cutting tool and a point on the workpiece is:

R
Twork actual = EV R Ttool actual (20)

where EV is the volumetric error homogenous transformation matrix representing position and orientation
errors between the cutting tool and workpiece. The position vector component of EV represents the
translations in the toolpoint`s coordinate frame that must be made to the toolpoint in order to be at the proper
location on the workpiece.
For implementation of error corection algorithms on machines with revolute and translational axes, in
order to move the axes to desired location specified by (20), inverse kinematic solutions should be used. For
a machine tool with translation axes only, volumetric errors can be calculated using the position vectors of
the HTM given in (7) through (11).
The error corection vector with respect to the reference coordinate frame can be obtained from the
following matrix equation:

R R R
Px Px Px
R
Pcorrection = Py = P
y
P
y (21)
Pz correction Pz tool Pz work
 Doina MARIN 106

R
Pcorrection = R Ptool R Pwork = P R T3 3Ttool ( ) actual
(
P R T1 1T2 2Twork ) actual
(22)

Now, we can calculate Px, Py and Pz , the volumetric error compensation components in the X, Y and Z
directions. The expanded form of (22) is obtained as follows:

Px = T x T y z 3 ( z ) + T z y 3 ( z ) + x 3 ( z ) + a 3
( )
1 z1 ( y ) z 2 ( x) y1 ( y ) y 2 ( x) (Tx + a 3 x a 2 a1 )
( ( x) ( y ) + ( y ) ( x) )(T + b b y b )
z2 z1 y1 x2 y 3 2 1 (23)
( ( x) + ( y ) ( x) + ( y ) )(T + z + c c c )
y2 z1 x2 y1 z 3 2 1

x ( x) a + ( y )( ( x) + b ) ( y )( ( x) + c )
x2 2 z1 y2 2 y1 z2 2 x1 ( y ) a1

Py = T y + Tx z 3 ( z ) Tz x 3 ( z ) + y 3 ( z ) + b3
( )
z1 ( y ) + z 2 ( x) + x1 ( y ) y 2 ( x) (T x + a 3 x a 2 a1 )
(
( z 2 ( x) z1 ( y ) x1 ( y ) x 2 ( x) ) T y + b3 b2 y 1 ) (24)
( )
y 2 ( x) z1 ( y ) x 2 ( x) x1 ( y ) (Tz + z + c 3 c 2 c1 )
z1 ( y )(x + x 2 ( x) + a 2 ) y 2 ( x) b2 + x1 ( y )( z 2 ( x) + c 2 ) y y1 ( y ) b1

Pz = Tz + T y x 3 ( z ) Tx y 3 ( z ) + z 3 ( z ) + z + c 3
( )
x1 ( y ) z 2 ( x) + y1 ( y ) y 2 ( x) (Tx + a 3 x a 2 a1 )
( ( x) +
x2 x1 ( y ) + y1 ( y ) z 2 ( x) )(T y + b3 b2 y b1 ) (25)
( ( y ) y1 y 2 ( x) )
x x1 ( y ) x 2 ( x) + 1 (Tz + z + c3 c 2 c1 )
(
+ y1 ( y )(x + x 2 ( x) + a 2 ) x1 ( y ) y 2 ( x) + b2 z 2 ( x) c 2 z1 ( y ) c1)
The resultant volumetric error (ERV ) can be obtained by the following equations:

(
E RV = Px2 + Py2 + Pz2 )1/ 2
(26)

The measurement offsets of the axes origins, which occurred during laser measurement of the linear
and angular errors of the machine tool, are a, b and c and given in machine coordinate system.
The parametric errors of the machine are modelated by using statistical technicques such as multiple
regression analysis or Artificial Neural Networks. These models are then updated in real time using
temperature sensors strategically attached to machine structure. As a result, calculated volumetric error
compensation values are downloaded as error compensation table to the machine controller to apply software
based error compensation..
In the application of CNC machine tools, the desired positions of cutter interpolated from the cutting
trajectory can be directly obtained from the NC program posted by CAD/CAM system. In order to execute
efficiently error compensation, a NC code was developed to identify and extract the nominal positions of
cutter for processing error prediction and compensation.
When the error compensated positions are determined, the software will automatically generate a new
NC program adopting the error compensated positions for accurate machining. In fig. 3 is represented the
compensation scheme for error of CNC machine tool.
107 Error prediction and compensation for cnc machine tools

Desired tool position Error compensated position

cutter
Identify position
desire Compensation Rewrite CNC Machine
Load NC cutter software NC codes controller Servo
Codes system
position

Fig.3 Error compensation scheme for CNC machine tools

5. CONCLUSION

Each slide of the three axis machine tool has been modeled using rigid body kinematics with small
angle approximation of angular errors and homogenous transformation matrix.
Volumetric error homogenous transformation matrix representing position and orientation errors
between the cutting tool and workpiece has been derived.
Volumetric error compensation components were derived using position vector of the HTM.
Using the derived volumetric error compensation component models and measured parametric
errors, the volumetric error compensation components in the X, Y and Z axis directions were calculated.
The volumetric error compensation values introduced in the controller of the machine tool can
correct the path error.

REFERENCES

1. Anjanappa, M., Anand, D.K., Kirk, J.A., Shyam, S., Error Corection Methodologies and Control Strategies for Numerical
Control Machining, Control Methods for Manufacturing Processes, ASME. DSC 7, pp.41-49, 1988.
2. Donmez, M.A., Blomquist, D.S., Hocken, R.J., Liu, C.R., Barash, M.M., A General Methodology for Machine Tool Accuracy
Enhancement by Error Compensation, Precision Engineering 8 (4), pp.187-196, 1986.
3. Duffie, N.A., Bollinger, J.G., Generation of Parametric Kinematic Error Correction Functions from Volumetric Error
Measurement, Annals of CIRP 34 (1), pp.259-262, 1985.
4. Kiridena, V.S.B., Ferreira, P.M., Kinematic Modeling of Quasistatic Errors of Three-Axis Machine Centers, International
Journal of Machine Tools and Manufacture 34, pp.85-100, 1994a.
5. Kiridena, V.S.B., Ferreira, P.M., Parameter Estimation and Model Verification of 1st Order Quasistatic Error Model for
Machine Centers, International Journal of Machine Tools and Manufacture 34, pp101-125, 1994b.
6. Kiridena, V.S.B., Ferreira, P.M., Computational Approaches to Compensating Quasistatic Errors of Three-Axis Machining,
International Journal of Machine Tools and Manufacture 34, pp127-145, 1994c.
7. Mou, J., Liu, C.R., A Method for Enhancing the Accuracy of CNC Machine Tools for on-Machine Inspection, Journal of
Manufacturing Systems 11 (4), pp.229-237, 1992.
8. Ni, J., Wu, S.M., A New On-line Measurement System for Machine Tools Geometric Errors, NAMRC Conference Proceedings
15, pp. 573-577, 1987.
9. Okafor, A.C., Ertekin, Y.M., Derivation of Machine Tool Error Models and Error Compensation Procedure for Three Axes
mertical Machining Center Using Rigid Body Kinematics, International Journal of Machine Tools and Manufacture 40, pp.1199
- 1213, 2000.
10. Wang, S.M., Ehmann, K.F., Measurement Methods for Position Error of a Multi-axis Machine Part I, International Journal of
Machine Tools and Manufacturing 39, pp. 951-964, 1999.
11. Wang, S.M., Ehmann, K.F., Measurement Methods for Position Error of a Multi-axis Machine Part II, International Journal
of Machine Tools and Manufacturing 39, pp.1485-1505b, 1999.
12. Zhang, G., et al., Errors Compensation of Coordinate Measuring Machines, Annals of CIRP 34 (1), pp.445-448, 1985.