Selection of Methods of Surface Texture Characterisation for Reduction of the Frequency-Based Errors in the Measurement and Data Analysis Processes.

Przemysław Podulka

doi:10.3390/s22030791

Abstract

Processes of surface texture characterisation can be roughly divided into measurement issues and analysis of the results obtained. Both actions can be fraught with various errors, some of which can be analysed with frequency performance. In this paper, various types of surface topographies were studied, e.g., cylinder liners after the plateau-honing process, plateau-honed liners with additionally burnished dimples of various sizes (width and depth), turned, milled, ground, laser-textured, ceramic, composite and some general isotropic topographies, respectively. They were measured with a stylus or via optical (white light interferometry) methods. They were analysed with frequency-based methods, proposed in often applied measuring equipment, e.g., power spectral density, autocorrelation function and spectral analysis. All of the methods were supported by regular (commonly used) algorithms, or filters with (robust) Gaussian, median, spline or Fast Fourier Transform performance, respectively. The main purpose of the paper was to use regular techniques for the improvement of detection and reduction processes regarding the influence of high-frequency noise on the results of surface texture measurements. It was found that for selected types of surface textures, profile (2D) analysis gave more confidential results than areal (3D) characterisation. It was therefore suggested to detect and remove frequency-defined errors with a multi-threaded performance application. In the end, some guidance on how to use regular methods in the analysis of selected types of surface topographies following the reduction of both measurement (high-frequency noise) and data analysis errors was required.

Highlights

Surface texture can be classified as a fundamental issue for characterising the properties of the manufactured parts that can support the process of control
It was found in previous studies that the detection of high-frequency measurement errors can be improved with the graph performance of power spectral density (PSD), auto correlation function (ACF) and frequency spectrum (FS)
The PSD enabled the derivation of the surface roughness and provided useful information on characteristic features which compose the microstructure of the films [36] or, for optical thin films [37]

Summary

Introduction

Surface texture can be classified as a fundamental issue for characterising the properties of the manufactured parts that can support the process of control. The whole process of surface texture evaluation [11], including the measurement and analysis of the results obtained, can be fraught with many issues that influence the accuracy of the studies performed. All of those errors can be roughly divided into those typical to the measuring method [12] and those caused by the digitisation or data processing [13], software [14], measuring object [15] or other errors [16,17]. One of the types of errors is facilitated when the measurement process occurs. The errors that occur while the measurement process takes place are often defined as noise [18]. From all of the studied types of measurement noises, e.g., instrument or instrument white [19], random [20], phase [21], signal-to-ratio [22] or, the measurement noise [23], the high-frequency domain (high-frequency measurement noise [24]) is most commonly studied in recent research

Objectives

Methods

Results

Conclusion