Featured Product
This Week in Quality Digest Live
CMSC Features
David L. Linville, Yongwoo Park, Nay Lin, and Yuanqun Lin
Overcoming the challenge of temperature control
Mike Richman
What if the Metre Convention had never happened?
Belinda Jones
Among the greatest attractions at the CMSC
Belinda Jones
Veteran metrologist Michael Raphael and former astronaut Dr. Bonnie Dunbar headline
Scott Sandwith
35th anniversary conference coming up July 22–26 in Orlando, FL

More Features

CMSC News
New NIST chip hints at quantum sensors of the future
Deal provides new and better solutions for customers
New portable, wireless optical coordinate measuring system
Full-field, high-accuracy noncontact part measurements with fully automated scanning
“Simplifying the lives of our customers while keeping them fully in control”
State-of-the-art facility built with Hexagon technology
ATS600 is the world’s first metrology-grade scanning laser tracker
Optimized for manufacturing, assembly, and construction workflows
Scott Knoche appointed to new position

More News

E. Kushnir, R. Karadayi, W. Clark, A. C. Affer, A. Naga

E. Kushnir, R. Karadayi, W. Clark, A. C. Affer, A. Naga’s default image

CMSC

Enhancing the Accuracy and Productivity of Super-Precision Turning Machining Centers

Published: Thursday, December 13, 2018 - 12:06

Modern super-precision turning centers must provide accuracy in the lower range of diamond-turning machines (cylindricity of less than 1–2 μm) and high productivity. Traditionally, diamond-turning machines have not been able to provide this because of their design limitations. However, high accuracy can be achieved if compensation of error is computed based on feedback from the results of measuring part dimensions and profiles.

There are systems that provide dimensional and sometimes profile feedback from the results of measurements performed outside the cutting envelope of a turning center (i.e., a lathe). For super-precision turning centers with relative short-duty cycles, this approach is not effective. Super-precision of a turning center offers the advantage of providing guideways to support the measuring device, as well as the performance of in-envelope measurement of part profiles and dimensions. Information obtained during this measurement can be used to compensate for errors that are correlated to variation in cutting force as a function of tool wear, as well as variation of parts-material characteristics in the batch.

This type of compensation makes sense only when increasing turning-center accuracy by mechanical means would be too expensive or not achievable. Using an envelope measuring system for error compensation may be justified only if the machining center feed system can perform small steps that are at least 0.1–0.2 μm, that is, 5–10 times less than what is required for profile accuracy in the range of 1–2 μm.

Profile representation for compensation procedure

For CNCs, the part profile is represented by data saved in the CAD/CAM system as a set of standard features or points representing the part profile and set of CNC-program commands for machining. The CNC program develops a tool pass along the part profile that is transferred to CNC feed servers to perform the required motion in the machine tool coordinate system. This system is represented as a grid with a defined step size. The minimum size of the grid is defined by CNC performance and is usually equal to 0.0001 mm or 0.00001 in.

In modern CNCs, additional functions include position and straightness error compensation. In error compensation applications, a grid step may be defined by the CNC programmer as a function of part size, required accuracy, and the available memory that is allocated for additional functions performed by the CNC. Usually up to 500–1,000 points for the axis are allocated for position and straightness compensation. Modern CNCs perform straightness compensation movements in a direction normal to the coordinate axis and only in points that belong to its compensation grid (see grid in figure 1). Position compensation is performed in the same points but along the coordinate axis.

It is obvious that the accuracy of profile compensation will be affected by the compensation grid step and, consequently, by the number of points involved. From another point of view, the required grid step is a function of profile waviness that will define the required compensation accuracy. In any case, the grid step used for compensation will be bigger than the minimum size of the grid used by the CNC during machining. This means that part representation for profile-error compensation must be a subset of the points used in representing the CNC. Because CNC profile representation is obtained during machining from CAD/CAM data for the purpose of automatic error compensation, the same CAD/CAM data have to be transformed to represent the part profile through a set of points with coordinates that belong to the grid used for compensation—to set points in which the CNC will perform computed compensation movements. Thus, this compensation movement is based at the CNC position and straightness compensation functions, and it will not alter the original machining program.


Figure 1: Compensation and measurement point’s representation

As seen in figure 1, the requirements for a compensation representation of the part profile can be described by a combined set of two points (Zf, X) and (Z, Xf), where: Zf are coordinate points along the Z axis in which straightness compensation in X direction and position compensation in Z direction may be performed; and Xf are coordinate points along the X axis in which straightness compensation in Z direction and position compensation in X direction may be performed. Coordinates defined as Z and X are any coordinates along Z and X axis, respectively, that do not belong to the compensation set. With this approach, for any point in the profile representation, at least one of the coordinates belongs to the compensation grid.

The compensation value in each of these points must be computed based on the part profile measurements.

Measurement procedure

A profile measurement program with certain amounts of measurements points (optimized by accuracy and productivity criteria) can perform measurements in the subset of a chosen set of compensation points that represent the part profile. For example, for the profile presented in figure 1, a different number of measurement points were chosen to represent regions of the profile with different curvature and position relative to the axes.

In regions 1 and 4, which are at a small angle, or parallel, to the X axis, the measurement points were chosen based on the assumed length of error wave of 8 mm (X axis ball screw lead). To ensure proper compensation, the harmonic distance between measurement points has to be at least 4 mm.

In regions 3 and 5, which are at a small angle, or parallel, to the Z axis, measurements points were chosen based on an assumed length of error wave of 10 mm (Z axis ball screw lead). The distance between the measurement points must be at least 5 mm. In region 2, the distances between the measured points were chosen to be 5 mm along the curve itself.

After the measured points are defined by chosen criteria, they are moved to the closest points of type (Zf, X) or (Z, Xf) in the compensation set. The obtained subset of points is used to perform the profile measurement.

For any given profile, programs that develop the set of compensation points and subset of points for measurement must be run only once before machining the part.

Error compensation procedure

After cutting is performed, the machined profile has to be measured at chosen points. The results of measurement are used to compute the profile error values in the compensation set of points (see figure 2).


Figure 2: Modeling of part profile error compensation process

All obtained measurement error data (i.e., measurement points of types (Zf, X) and (Z, Xf)) are used for error approximation of the compensation points not included in the measured subset in both directions. In this case even if, for some points, the measured error value in one of direction will not be used for compensation, it is still used for error approximation. Thus, all profile measurement data are used in computing to increase the calculation accuracy.

The compensation values have to be computed in a direction normal to the part profile. Projections of the computed normal error (i.e., the error vector along what is normal to the profile) at the Z and X axes are used for compensation. This approach allows compensation of the part profile, in the direction normal to the profile and not tangential to it. In addition, this approach resolves the issue with the value of position compensation in different points along parts of the profile that are parallel to the axes. This value will be equal to zero by definition.

This means that in region 4, only straightness compensation will be performed in the Z direction for compensation points with coordinates (Z, Xf) (as seen in figure 1). In region 5, only straightness compensation will be performed in the X direction for compensation points with coordinates (Zf, X).

Profile error compensation methodology simulation

Simulation of the proposed profile error compensation methodology was performed in Excel. The profile used for the simulation is shown in figure 1. The compensation grid at figure 2 has a step of 0.5 mm. At figure 2, this profile is represented as the sum of two sets of points (Zf, X) or (Z, Xf), with chosen points of the profile positioned at Z or X gridlines (as seen in figure 3).


Figure 3: Profile represented as the sum of two sets of points (Zf, X) or (Z, Xf), with chosen points of the profile positioned at Z or X gridlines

The first cut of the part profile was simulated by adding random errors to the profile points chosen for measurement. The part profile with errors are shown in figure 2 (curve 1 in figure 2 represents the original profile that must be machined; points “∆” are the results of measurements of the chosen points and curve 2, and an approximation of the profile errors).

The machining process was simulated by subtracting the computed compensation values from the profile that represent an approximation of profile errors (curve 2 in figure 2). Curve 3 in figure 2 represents the final profile obtained after part machining was simulated. As can be seen, the compensation process allows for a substantial decrease in the initial profile errors.

Requirements to enhance super-precision turning machining centers

The enhanced super-precision turning machining center that can machine hard materials (RC 55 and higher) in the range of turning accuracy of 1–2 μm has to include:
• Machine tool with high stiffness of the cutting loop (the variation in cutting force from pass to pass develops displacement in the cutting zone that are 5–10 times less than required accuracy)
• Machine tool with slides that can perform very small steps (less than 0.1 μm) with high repeatability of motion
• Measurement equipment that can perform measurements with an accuracy that is 5–10 times better than the required accuracy
• Measurement program that allow to perform measurement in defined points along part profile, developed by machine CNC control
• Compensation methodology that does not require cutting program modification
• Compensation program that will compute compensation values in chosen statistically defined grid points (compensation points) of the machine tool coordinate system
• Interfaces between the measurement program, compensation program, and the CNC that prevent any interrupted flow of data between the different components of the machine tool and the compensation system, and without operator intervention

The turning center model T51, as a representative of the family of super-precision turning machining centers (models T42, T51, T65—see figure 4), satisfies all the above-mentioned requirements. This machine has the ability to perform small steps of 0.1 μm and has a very high level of repeatability, as it is shown in figure 5.

Machine accuracy was checked by cutting an aluminum cylinder with a 75 mm radius and length of 250 mm. The machine had a straightness in the cutting envelop of less than 0.2 μm, and perpendicularity of less than 0.2 μm/m, as shown in figure 6.


Figure 4: Super-precision turning centers by Hardinge


Figure 5: Small step test data for X and Z axis slide (T51 machine)


Figure 6: Comparison of straightness of aluminum cylinder measured by Roundcom and by a probe mounted at the T42 machine (need all data to prove that data are close in 0.2 μm range).

The T51 turning center is equipped with a FANUC I-31 CNC. The FANUC CNC has the ability to perform compensation in grid points. Between compensation of the grid points, the FANUC CNC performs linear interpolation of errors. The order of interpolation used to compute the compensation values in the grid compensation points, based on data obtained in the measured points, must be defined based on the harmonic components found in the errors of the machined profile. In the following example, the linear interpolation was used to compute the profile error compensation values in the compensation grid points.

Measured equipment

A measurement sensor is set up at the same slides as the cutting tool. The accuracy of this sensor is in the range of 0.25 μm. The measurement of the machined profile is performed when the slide is moving along different parts of the guideways than during cutting. For this reason, the measurement motion must be correlated with the cutting motion to deliver adequate data. This correlation was performed by measuring the control profile in the X and Z directions, and by electronic compensation of the measurement pass errors so that the results of the measurement profile at Roundcom and at the machine tool are within an accuracy of 0.2 μm, as shown in figure 6.

Measurement and error calculation program

The software needed to represent part profile in grid points, and to perform the required measurements and calculations, was developed as an extension of the AAT3D CappsNC In-Process Measurement Program. A FOCUS support interface of the FANUC CNC was used to ensure a continuous flow of data between the CNC, CappsNC, and its extension at all steps of the profile measurements, analyses, and cutting.

Verification of developed profile error compensation methodology

The developed compensation procedure and developed software were verified by cutting a part of a sphere at the T51 super-precision machining center. The verification was performed in three steps.

During the first step, a standard CAM program was developed to cut a nominal profile, presented at figure 7 (curve 1), by adjusting the program parameters that define the profile. The part profile was measured after every cutting, and corrections in the part radius, part center position, and insert radius were performed to obtain the required profile. It took a substantial amount of cutting and measurement time to get the profile close to nominal (see curve 2 in figure 7; the radial error scale is equal to 1,000).

During step two, artificial errors were introduced that were obtained during the CNC programming in step one, with the goal of producing a substantial error in the part profile during machining. The result of measuring this profile is presented in figure 7 by curve 3 (error scale = 1,000).

Data obtained when measuring othe profile with artificial error were used to compute the compensation values in the grid points and applied while running the CAM program used during step two of the verification process. The profile obtained at this step is presented in figure 7 by curve 4 (error scale = 1,000).

Profile errors of all three cut cuts are presented also in figure 8 in original values.

It follows from the presented data that the proposed part-profile error-compensation methodology can be applied to increase the accuracy of machining and reduce program development and setup time.


Figure 7


Figure 8

Discuss

About The Author

E. Kushnir, R. Karadayi, W. Clark, A. C. Affer, A. Naga’s default image

E. Kushnir, R. Karadayi, W. Clark, A. C. Affer, A. Naga

E. Kushnir and W. Clark are with Hardinge Inc. of Elmira, NY; R. Karadayi, A. C. Affer, and A. Naga are with Applied Automation Technologies of Rochester Hills, MI.