77141839-filters-outliers.pdf

Calypso

advanced techniques

Filters, Outliers and the Scanning CMM Introduction This document will describe the use of Filters, Outliers and Evaluation Methods for the Scanning CMM. Although Calypso has been used for all instructions, plots and output, the techniques described here apply to UMESS as well. Also, it must be understood, the following techniques apply to the Scanning CMM. Actually not only is it “advisable” to use scanned features but actually impossible to apply filters to features with low data density. In addition to data density, other areas of discussion will be: •

Application of Filters and outliers within the Calypso measurement plan.

•

Why and when to apply filters and outliers.

•

What’s the Standard?

•

Error messages.

•

Evaluation Methods.

•

Centers for Out-of-Roundness.

•

Instrument Response. Cutoff, UPR and Lambda c.

•

Filter types, Gauss and 2-RC.

•

Low and High pass filters.

•

Connect Segments.

•

Elimination of Outliers.

•

Range of Data Reduction.

The information in this document is from a variety of resources including the Calypso manual and on-line help.

Defining Filters and Outlier Elimination There are several different points in Calypso at which you can parameterize and activate filtration and outlier elimination: – For an individual characteristic – For an individual feature – As defaults for the characteristic groups and the references and alignment elements of the coordinate systems. The setting for the characteristic always takes priority. Iffiltration/outlier elimination is not activated for the characteristic, the setting for the feature applies. If filtration/outlier elimination is not activated for the feature, Calypso refers to the default settings for the characteristic groups. A feature obtained by recall from a feature filtered beforehand is automatically not filtered subsequently, even if filtration is activated.

Default Settings To set the Default Filter and Outlier settings, select Resources > Filter/Outlier Elimination.

Settings at this level are not applied to existing characteristics, only those subsequently created. Features containing filter and/or outliers settings will override these defaults.

Feature Settings The next level at which Filters and Outliers can be set is within the Feature. Open the feature window and select Evaluation. These settings will be automatically applied for each characteristic in which the feature is used. This overrides the default settings.

Characteristic Settings From within the characteristic select the feature button.

From this widow you can override settings from either the feature or default. Also it is possible to unselect settings applied by the feature or default.

Why Use Filters From the Calypso on-line help: Filters can segregate the waviness profile of a geometric feature from the effects of surface roughness. Filters, therefore, are of assistance in maximizing measuring accuracy. The filter methods supported are Gaussian (ISO 11562) and 2 RC (ISO 4291), while the filter types are lowpass and highpass. The new statistical data of the geometric feature are obtained from the filtered measuring data after compensation.

What’s the Standard? Often filters are specified on the print or inspection plan. In some cases, roundness for instance, a filter might be implied by the standard: From ANSI B89.3.1 – 1972 Measurement of Out-Of-Roundness: section 3.2 ...If complete measurement conditions have not been specified the: Method of Assessment—Minimal Radial Separation Instrument Response—50 Cycles per Revolution

.02

In other words: Implies:

.02

MZC

50

That is: this surface shall be round within .02mm as assessed by the MZC method with 50 cycles per revolution response. If the print is specific regarding the filter it might be shown as thus:

.02

LSC

150

That is: this surface shall be round within .02mm as assessed by the LSC (Gauss) method with 150 cycles per revolution response.

Standards are available on the internet at: global.ihs.com

Using Filters in Calypso:

.02

MZC

50

The above feature control frame requires the evaluation of Roundness (also called Circularity or Out-Of-Roundness)

Data Density: It is important to understand that the use Filters and Outliers are really best applied to scanned features. From the Calypso on-line help: It is advisable to restrict the use of filters to instances in which the number of points is high, in other words they should be used only for scanned features. All features measured using scanning methods can be filtered. Calypso also requires a minimum of 5 (soon to be changed to 7) points per Lambda c (cutoff) or Undulations Per Revolution (UPR) frequency. Use of insufficient data density will result in the filter not being applied and a message printed in the default printout. If for instance you choose to evaluate roundness with 100 data points and a UPR of 50, no filter will be applied, and a message will be reported in the default printout: Incorrect filter value (Lambda_c or UPR) Increasing the data to a range of 101 to 250 points and once again no filter will be applied. The default printout message: Filtering with less than 5 points per undulation. Increasing the number of points to 251 or greater and a filter will be applied. A message reporting the filter type and cutoff will be reported in the default printout: Low-pass (Form):Gauss Undulations Per Revolution:50

Other Messages •

The Compact Protocol shows when a filter has been applied, but no error messages.

•

If a filter is selected but not applied, the plot window shows “none”, otherwise it reports the filter being used.

•

The Custom Printout never shows filter messages.

Setting Filters in Calypso

.02

MZC

50

Select the Roundness Characteristic and scan the required feature. Check the box to apply the filter. Notice the evaluation method is shown next to the check box. In this case the default setting matches the Standard. That is, Minimum Zone Center with a Gauss filter at 50 UPR. Of course, all of these settings are user definable. They can be specified by the operator either here, in the characteristic, within the feature or as a Calypso default filter setting. The above feature control frame specifies the method to be used for the evaluation is a Minimum Zone Center (MZC). Let’s look at the other possibilities.

Evaluation Methods Click on the Evaluation Method Parameters button to open the Evaluation Methods window. This displays several selections: • LSQ Element (Standard)… aka, Gauss. • Minimum-Element… aka, MZC • Minimum Circumscribed Element • Maximum Inscribed Element • Inner Tangential Element • Outer Tangential Element It is of interest here to consider the use of the word “Element” in this dialog window. For example, one might think that the selection Maximum Inscribed Element might better be called Maximum Inscribed Circle. However, Calypso can apply filters and outliers to a variety of features beyond the circle we are considering here. From the Calypso on-line help: Individual measuring points of the geometric features 2d straight, plane, circle, cone, cylinder and sphere can be purged from the measured values as Outliers.

Centers for Out-of-Roundness Measurement (4 possibilities) The centers of the measured polar profile which may be used to determine the Out-of-Roundness value when specified are those related to one of the following alternative methods of Out-of-Roundness assessment: 2.8.1 Minimum Radial Separation (, MRS, Minimum Zone Center, MZC, Tschebyscheff) This center is that for which the radial difference between two concentric circles which just contain the measured polar profile is minimum. This is also known as the center for minimal Total Indicator Reading (TIR). The British Standards Institution publication 3730:1964 refers to it as Minimum Zone Center (MZC).

Minimum Radial Separation (, MRS, Minimum Zone Center, MZC, Tschebyscheff)

2.8.2 Least Squares Center (LSC, Gauss) This Center is that of from which the sum of the squares of the radial ordinates of the measured polar profile has a minimum value.

2.8.3 Maximim Inscribed Circle (MIC) This center is that of the largest circle which can be inscribed within the measured polar profile. This is also known as the plug gage center and is generally used for internal diameters.

Least Squares Center (LSC, Gauss)

2.8.4 Minimum Circumscribed Circle (MCC) This center is that of the smallest circle which will just contain the measured profile. This is also known as the ring gage center and is generally used for external diameters.

2.9 Preferred Center The center from which the out- of-roundness value shall be determined unless specified otherwise is the Minimal Radial Separation Center.

Maximim Inscribed Circle (MIC)

In Calypso the MZC or Tschebyscheff is correctly used by default for Roundness evaluation. Other centers are user selectable. In Calypso versions 3.0 and higher, within the characteristics feature selection window, there is a button for Evaluation Parameters.

Minimum Circumscribed Circle (MCC)

Instrument Response (Cutoff) Cycles Per Revolution (Undulations Per Revolution, UPR) If all of the radial deviations of a circular cross section were fully and completely represented by a measured profile, the presence of high frequency surface irregularities could mask any lobing conditions or the form of the profile. In the ANSI standard the term used is Cycles Per Revolution Response, in Calypso it is called Undulation Per Revolution. A UPR filter of 50 means that the measured profile has been attenuated by a filter which has reduced by a percentage (Gauss 50%, 2RC 75%) the amplitude of the sinusoidal lobing which occurred at a regular interval of 50 lobes per revolution. Note that the 50 UPR frequency is base on angular displacement rather than time. It might be expressed that the cut-off of 50UPR represents 50 segments of the circle. Reducing the number of UPR will tend to smooth out small scale irregularities while higher frequency are more inclusive of total surface texture. At least 5 measuring points are required per undulation.

Wavelength Lc (Lambda cutoff) Profiles at three different UPR settings

Flatness and straightness are evaluated with a cut-off describe by a linier dimension. The default for Calypso is 2.5mm. A Wavelength Lc of 2.5 means that the measured profile has been attenuated by a filter which has reduced by a percentage (Gauss 50%, 2RC 75%) the amplitude of the sinusoidal profile which occurred at a regular interval of 2.5mm.

Filter Methods: Gauss The Gaussian filter has been adapted as the industry standard for the measurement of roundness and surface geometry. It is a digital, phase corrected filter. Its quick frequency response allows the accurate assessment of the surface profile. The Gaussian filter is defined to have 50% transmission at the cutoff wavelength. In other words; a measured sine with 50 UPR and an amplitude of 1.0 has after filtering with gauss filter (50 UPR) an amplitude of 0.5.

2-RC-Filter (2-CR) The 2RC (or 2CR) filter is an older standard filter used in surface roughness measurements. It Is typically implemented as an analog electrical filter, 2 RC filters in series, separated by a buffer. It Is not phase-correct, and it has a frequency transmission of 75% at the cutoff wavelength. The 75% level was apparently chosen because the 2RC filter has a long “tail” in its frequency response, and, therefore, much of the long wavelength components above the 50% cutoff remains in the roughness. The cutoff at 75% more accurately retains the intuitive sense of being the wavelength boundary between roughness and waviness.

2RC

Gauss

Some refer to the above filter as a 2CR filter because the capacitor comes before the resistor in each pair. However, electrical engineers refer to either order as an RC filter and distinguish between the two by calling one a high pass filter and one a low pass filter. “Pass” refers to what frequencies make it through the filter. Comparrison of Gauss vs. 2RC frequency responce

Both the Gauss and 2RC filters separate roughness from the form error.

Surface Profile

Waviness Form

Lowpass, Highpass Filter Types You may select a Lowpass or a Highpass filter. The Calypso default is Lowpass. A Lowpass filter type reduces the short-wave surface phenomena (High frequency) and does not disrupt the long-wave (Low frequency) effects of waviness and geometric deviation.

Caution Should be Used!! Caution should be used when changing from the default Lowpass to Highpass filter. Selecting the Highpass filter will remove all waviness (what you’re trying to evaluate with roundness, flatness, straightness, etc.) leaving only the high frequency surface roughness. Rather than using a CMM, surface texture might be better evaluated with a dedicated surface finish gage such as those in our TSK product line. That said, an experienced metrologist might glean insight into a process with the cautious use of the Highpass filter. In the instance where the measured feature has extremely poor geometry the low frequency deviation will prevent the use of a adequate magnification factor for plotting. Removing the low frequency form deviation will allow a much higher magnification, possibly revealing, machining, tool path or other process problems.

Example of a Out-of-Round feature with a form error of 4.8mm. A Lowpass plot must be reduced to a magnification of two or it’s off the chart. This obscures all high frequency form deviations.

However, as this custom printout shows, care must be taken. Two evaluations of the same feature produce two extremely different results!!

By selecting a Highpass the magnification can be increased to 400 revealing tool marks and chatter.

Connect Segments From time to time it may be necessary to scan a feature with interrupted paths. Possibly keyways inside a bore. Because of the limitations of 5 data points per cutoff (UPR or Wavelength Lc) filtering might not work. If you activate this checkbox the segments of an interrupted contour are connected to form a continuous contour during filtration with as many points as possible being taken into account. This means that you can select a stronger filter.

Eliminating Outliers A measured point is tagged as an outlier if it is further than a defined threshold (threshold = factor * standard deviation) from the computed Gaussian element. Outliers are geometric points that differ significantly from the geometric form yielded by the other measured points and as such, they can produce a large error when the compensatory element is calculated. An error of this nature easily propagates through the actual-value determination of the corresponding characteristic.

Why is Outlier Elimination Necessary? If, for instance, you are checking roundness on a dedicated form tester, you are probably in a inspection lab doing the work. If the characteristic is out of tolerance the inspector might clean and recheck the feature. Often, with CMM’s, there may not be this opportunity. With outlier elimination, and user definable parameters, the feature data can be “cleaned” by the software. Additionally, you do not want to make a process change based on a piece of dirt or imperfection in part material.

Selecting Outlier Elimination In Calypso As with filters there are several different points in Calypso at which you can parameterize and activate outlier elimination: – For an individual characteristic. – For an individual feature. – As defaults for the characteristic groups. At each of these points the Outlier Mode dialog window can be accessed.

Factor For Outlier Sets the outlier factor (factor * standard deviation) for outlier elimination. This factor can be set, individually, for inside and outside the material.

Range of Data Reduction Only Outlier – Removes only the outlier then recalculates the feature. Additional outliers may be eliminated in subsequent iterations. Include Adjacent Points – Operator can determine and number of points on each side of the outlier to also be eliminated. To Computed Feature – Eliminates the outlier and all point to the computed (Gaussian) feature. Outlier Elimination is reported on the Default Printout, Custom Protocol (see next page) and graphically represented on plots by the absence of deviation lines (below).

Calypso Compact Protocal - Range of Data Reduction example One plane with three evaluations. See bold print. ============================================================================================== C A R L Z E I S S / C A L Y P S O 3.2.-beta30 COMPACT PROTOCOL -1——————————————————————————————————————————————— Measurement Plan Operator Date Part Number Filters and Outliers Master July 21, 2001 32 CONTURA ——————————————————————————————————————————————— Names Description Actual Nominal Utol Ltol Deviat. Histogr. ============================================================================================== Plane4(Only Outlier) Minimum Zone Plane #P (222) Min = (3) -0.0030 Max = (57) Z 0.0373 0.1000 X -83.5389 -82.5000 Y 4.8318 5.0000 A1- X/Z -0.0171 0.0000 A2- Y/Z -0.1664 -0.1000 Outlier Elimination Inside Workpiece : 2 Outside Workpiece : 2 No Filter Only Outlier GDT Flat 0.0100 S =

0.0020

0.0030

Form =

0.0061

0.0061

|—

______________________________________________________________________________________________ Plane4(Outlier & Adjacent Points 4) Minimum Zone Plane #P (217) S = 0.0018 Min = (3) -0.0029 Max = (57) 0.0029 Form = 0.0057 Z 0.0376 0.1000 X -83.5389 -82.5000 Y 4.8317 5.0000 A1- X/Z -0.0169 0.0000 A2- Y/Z -0.1580 -0.1000 Outlier Elimination Inside Workpiece : 0 Outside Workpiece : 9 No Filter AP4 Outlier GDT Flat 0.0100 0.0057 |— ______________________________________________________________________________________________ Plane4(Outlier to Computed Feature) Minimum Zone Plane #P (217) S = 0.0015 Min = (3) -0.0026 Max = (57) 0.0026 Form = 0.0051 Z 0.0381 0.1000 X -83.5390 -82.5000 Y 4.8315 5.0000 A1- X/Z -0.0166 0.0000 A2- Y/Z -0.1443 -0.1000 Outlier Elimination Inside Workpiece : 4 Outside Workpiece : 5 No Filter CF Outlier GDT Flat 0.0100 0.0051 |— ______________________________________________________________________________________________