Elsevier

Precision Engineering

Volume 38, Issue 2, April 2014, Pages 439-442
Precision Engineering

Technical note
Periodic Gaussian filter according to ISO 16610-21 for closed profiles

https://doi.org/10.1016/j.precisioneng.2013.09.004Get rights and content

Highlights

  • The periodic Gaussian weighting function for closed profiles is introduced.

  • The weighting function enables the filtration of arbitrary measuring lengths.

  • A comparison to the periodic spline filter is given.

  • A Matlab script for the digital implementation is available.

Abstract

It is well known that the Gaussian filter according to ISO 16610, part 21 has nice metrological properties but also has some restrictions. The evaluable measuring length of open profiles is shortened by so-called filter running-in and running-out lengths which are a factor of the width of the Gaussian weighting function. ISO 16610-21 suggests a simplistic transfer of filtering of open profiles to filtering of closed profiles (e.g. roundness measurements). And this leads, due to the running-in and running-out lengths, to an unnecessary restriction. Because in case of a closed profile, it has to be taken into consideration that the profile can be extended infinitely and thus any measuring length of an equivalent open profile can be reproduced. Hence, as this paper shows, the measuring length for closed profiles can be arbitrary for any given width of the Gaussian weighting function.

Introduction

The standard ISO 16610-21 defines the Gaussian filter for dimensional metrology [1]. Qiang-Qian Jiang as the project leader of ISO 16610-21 and Paul Scott as the convenor of working group 15 of ISO TC 213 (both are international experts for surface metrology and working at the University of Huddersfield) were significantly involved in writing the new standard. A nice guideline dealing with the implementation of the Gaussian filter was published by Michael Krystek [4], an internationally recognized expert for surface metrology as well.

It has to be mentioned that ISO 16610-21 serves as a fundamental basis for industrial and metrological applications. And it is well known that the Gaussian filter not only has nice metrological properties but also has some restrictions. For example, open profiles have a finite length. Hence, the evaluable measuring length is shortened by so-called filter running-in and running-out lengths which are a factor of the width of the Gaussian weighting function. When needed, these sections can be still used by applying ISO 16610, part 28 [3]. However, filtering is then no longer phase-correct and therefore these sections are also called the end effect regions of the filter. ISO 16610-21 suggests a simplistic transfer of filtering of open profiles to filtering of closed profiles (e.g. roundness measurements). But this leads to an unnecessary restriction. Because in case of a closed profile, it has to be taken into consideration that the profile can be extended infinitely and thus any measuring length of an equivalent open profile can be reproduced. The result is a periodic convolution of a closed profile of arbitrary length and a periodic weighting function derived from the Gaussian weighting function with any given width.

Section snippets

Linear profile filtering with the Gaussian filter

The basis of linear profile filtering is the convolution integral. The procedure can be seen as determination of a moving average with a weighting function. The following relation applies to an open, infinitely extended profile:w(x)=z(xξ)s(ξ)dξ=z(ξ)s(xξ)dξ,s(ξ)dξ=1where w(x) is the filter line, z(x) is the unfiltered profile and s(x) is the weighting function of the filter. An important requirement for defining the weighting function s(x) is that the area under the function

Linear periodic Gaussian filter for closed profiles

For closed profiles with a short measuring length (e.g. a cylindrical component with a small diameter), it may not be possible to use the “classical” Gaussian filter when the cut-off wavelength λc is a high value. However, this restriction is not necessary. A closed profile can also be handled as an open profile of infinite length. Then a periodic representation, with the circumference serving as the period length, is chosen. This means that the filter can also be used for geometries with small

Example

If a cosine-shaped profile with a wavelength of λ = 0.8 mm is measured over the measuring length L = 1.6 mm and a cut-off frequency of fc = 1 is assumed, the result is f/fc = 2. ISO 16610-21 does not recommend the evaluation of this profile for precision measurements. With the amplitude transfer from Eq. (1), a value ofa˜1a˜0=124=0.0625can be determined. When the periodic convolution integral of the cosine profile is calculated, it becomes clear that filtering is easy to accomplish and that the amplitude

Conclusion

Due to the required precision, ISO 16610-21 specifies restrictions for use of the Gaussian filter for roundness profiles with a short measuring length (e.g. cylindrical components with small diameters). These restrictions are assumed by simply applying the filtering process of open profiles to that of closed profiles. But this limitation is not necessary when the periodic Gaussian weighting function is considered. Then the filter can be used for small diameters and small values of the cut-off

References (4)

  • ISO 16610-21

    Geometrical Product Specifications (GPS)—Filtration—Part 21: Linear profile filters: Gaussian filters

    (2011)
  • ISO/TS 16610-22

    Geometrical Product Specifications (GPS)—Filtration—Part 22: Linear profile filters: Spline filters

    (2006)
There are more references available in the full text version of this article.

Cited by (5)

  • A form error evaluation method of honeycomb core surface

    2022, Measurement: Journal of the International Measurement Confederation
    Citation Excerpt :

    This problem is consistent with the idea of determining the baseline for surface topography evaluation. In ISO11562, the centerline by Gaussian filtering is taken as the baseline for dimensional metrology [25]. Since most data of the honeycomb cell wall locate at the boundary due to its thin wall structure, Gaussian filtering is not suitable owing to its significant boundary effect.

  • Description and validation of a circular padding method for linear roughness measurements of short data lengths

    2020, MethodsX
    Citation Excerpt :

    However, if the available space does not allow parallel tracks, a single track can be padded to itself, start-to-end, by “circular padding” (Fig. 9). Subsequently, the “circularly padded” profile can be unwrapped in a procedure similar to the ISO 16610–21:2011 [5] method of measuring roughness along the circumference of a cylindrical object [12]. This procedure conserves at both start and end a copy of the profile sufficiently long to maintain a sufficient measured length after filtering, i.e. at least λc on either side.

  • Filtering effects on involute gear form measurement

    2023, Measurement Science and Technology
View full text