826

826

TEXTILE RESEARCH JOURNAL

Characterization of Roughness–Friction: Example with Nonwovens STEPHANE FONTAINE,1 CYRIL MARSIQUET, MARC RENNER,

AND

MARIE-ANGE BUENO

Ecole Nationale Superieure des Industries Textiles de Mulhouse, University of Mulhouse, France

NATHALIE NICOLLETTI Ecole Superieure des Sciences Appliquees pour l’Ingenieur, University of Mulhouse, France ABSTRACT In several technical applications, it is necessary to accurately determine the surface state of materials. In order to do so, a new method to evaluate the surface state of materials has been developed. This method gives roughness–friction criteria, based on the principle of a “blade-disc” type tribometer, where the analyzed surface is the disc. In this study we have demonstrated the effectiveness of the method in a study of two types of nonwovens. The first type of nonwoven is intended for use for female or baby hygiene and the second type for baby skincare wipes. The nonwovens compared have surfaces with different structures and different compositions. In addition it is shown that the method is able to distinguish fine modifications of the surface state.

The baby wipes industry needs to compare and control softness parameters along the production process. One of the problems is to provide a measurement method, which is efficient and sensitive enough to control quality. This quality must be evaluated on nonwovens with very complex random surfaces. The surface measurement of nonwovens and random structures is not a new idea. Pourdeyhimi et al. [4, 5, 7] proposed a large study on fiber orientation measurements in nonwovens, using different optical methods and image analysis to track fibers in real and artificially produced images. First, the authors reported a “direct fiber tracking method that allows each fiber segment to be tracked ”. Images were analyzed, detecting the presence or the absence of a fiber. They have computed the orientation distribution function (ODF) that evaluates the frequency appearance of different orientation angles. A second evaluation method has been elaborated using “the power spectrum from a two dimensional Fourier analysis image” [4, 5, 7]. In this method, the structure is considered as a composition of “spatial details in the form of brightness transitions cycling from light to dark and from dark to light. The rate at which these transitions occur is the spatial frequency linked to the fiber orientation”. A Fourier transform (FFT) decomposes an image from its spatial domain of intensities into a frequency domain. Magnitudes of the various frequencies are evaluated by gray scale levels on the transformed 1 To whom correspondence should be addressed: e-mail: s.fontaine @univ-mulhouse.fr

images and represent the rate at which intensity transition occurs in a given direction. Subsequently, the FFT enables the ODF to be evaluated on a nonwoven. Moreover, the “equivalent Pore concept” was used by Salvado [9] in order to quantify the ODF on spunbonded nonwovens designed for the baby diaper industry. This method computes the cumulative length of straight segments (fibers) in all the directions of the nonwoven (two-dimensional analysis only). When “reorganizing the obtained lengths according to their orientation and putting them end to end, the cumulative length draw a curve in the plane” [9]. This curve is a semicircle in the case of an isotropic distribution of the segments and a semi-ellipse in the case of most nonwovens. Other studies have been made to evaluate random surfaces. Pourdeyimi and Sobus [6, 8] have studied the surface intensity and roughness of carpet in order to quantify the appearance changes on carpet surfaces due to mechanical wear. Assuming that there are three categories of carpet structures, namely strongly ordered structures, weakly ordered structures and disordered structures, the authors have proposed that carpet surfaces can be characterized according to global and local parameters from image analysis. Global parameters are one-dimensional statistical parameters which quantify pile compression or overall fiber disorder and local parameters enable the pile direction to be quantified in terms of overall pile disorder, tuft density and tuft shape. These studies show that the characterization of random surface state is a very complex problem. Even if image analysis provides the opportunity to quantify fiber

Textile Res. J. 75(12), 826 – 832 (2005) DOI: 10.1177/0040517505057634

© 2005 SAGE Publications

www.sagepublications.com

DECEMBER 2005

827

distribution on nonwoven structures or appearance loss on carpets, these “non-contact” techniques cannot be sufficient to quantify the touch of such products. Moreover, for several years, textile products have become more and more complex and more research has been directed towards the study of the tactile properties of these structures. In order to simplify the problem, we propose to separate the characterization of textile handle in two different ways: first, it is possible to consider the touch when fingers are just moving on a simply laid surface. During this motion the hand is sensitive to friction, surface roughness, transversal compression and cool/warm feeling. The second aspect of handle is by touching the textile surfaces with full hand. During this action, fabricbending properties are interacting with the physical properties previously described. Textile products are made up of different levels of structures issuing from various processes and are composed of fibers made of different materials. The fabrication process induces the presence of hairiness on the studied surface. These hairs are surface fibers which escape from the cohesion steps during the manufacturing processes. Hairiness and structure are strongly linked since some structures are naturally more hairy than others. In 1995, Bueno et al [1, 3] proposed a method to measure the surface state of textiles. This method measures the friction and the roughness of a fabric. The sample is fixed on a rotating plate and a 0.5-mm-diameter cylindrical probe is rubbed on the fabric surface. An accelerometer is fixed to this probe and measures the vibrations due to its penetration into the fabric structure. A FFT of the temporal signal provides a spectrum knowing that an average of several spectra gives an autospectrum. In the autospectrum, the energy of the peaks depends on the magnitude of the structure relief. This tribometer was multi-directional and able to measure fine modifications of surface states using the periodicity properties of the tested structure. However, two major disadvantages persisted: first, it was only possible to measure variations of treatments on the same surface and second, random structures could not be measured. The present study therefore introduces a measurement method which provides solutions to the two previous problems and obtains “roughness–friction” [2] criteria of the tested surface.

Experimental MEASUREMENT METHOD The authors have developed a patented measurement method [2] to obtain “roughness–friction” criteria for a tested surface (Figure 1). The sample is clamped on a rotating carrier and a very thin metallic blade (50 ␮m

thick) rubs the tested surface. During the contact, the blade vibrates according to eigenvalues of frequencies (vibration modes). Strain gauges are fixed on this blade and measure its vibrations (eigenvalues). A Fourier analysis of the temporal signal from the sensor entails the computing of the power spectrum relative to frequency with the help of a spectrum analyzer. The power spectrum is PS( f ):

FIGURE 1. Measurement method.

PS共 f 兲 ⫽ 兩X共 f 兲兩 2

(1)

where f is the frequency (Hz), X( f ) is the Fourier transform of the temporal signal x(t), which corresponds to the signal from the sensor (volts). The power spectrum relative to frequency gives a spectrum as well as an average of several spectra computed during one or more rotations of the sample carrier (Figure 2). This is an autospectrum in which the magnitude of the peaks of each mode depends on the surface state of the tested samples. The frequencies of the vibration modes are experimentally determined owing to this autospectrum (center of the peak). Moreover, to extract the energy of these peaks (equation (2)), it is necessary to integrate PS( f ) between two frequencies f1 and f2 (gray zone in Figure 2). In order to simplify the problem, only the modes 1 and 3 were studied. The reason why mode 2 has been eliminated will be explained later. Energy关 f 1 ;f 2 兴 ⫽

冕

f2

兩X共 f 兲兩 2

(2)

f1

When rubbing on the tested surface, the sensor is excited by surface events such as friction and roughness where friction is linked to (micro-) asperities and adhesion which depend on thin contaminant films that pollute

828

TEXTILE RESEARCH JOURNAL PERTINENT MODES

FIGURE 2. Autospectrum analysis.

surfaces. Moreover, the roughness Ra of a profile can be defined as: Ra ⫽

1 L

冕

L

兩y共 x兲 ⫺ y៮ 兩 䡠 dx

(3)

0

where L is the considered length for the integration, y៮ is the average line of the profile and y(x) the altitude of a point positioned at an abscissa x (Figure 3).

The sensor was studied by means of modal analysis using the finite element method. The code used was CASTEM 2004. This code has been developed by the Commissariat a` l’Energie Atomique – CEA – DEN/ DM2S/SEMT (http://www-cast3m.cea.fr/cast3m/index. jsp). This code has been chosen because of its great flexibility and because of its international recognition. This code is presented as a standard one with a preprocessor, where a program in Gibiane Language must be written. This program includes geometry, limit conditions, materials definitions and the analysis type that is required (thermic, dynamic and static mechanics. . .). The model was the simulation of a clamped-supported pre-loaded blade with the size of our sensor. The aim of this study was to compute the eigenvalues of frequencies of the sensor and the mode shapes for each studied frequency. Only the modes 1, 2 and 3 were studied because of the limited signal/noise ratio for modes of frequencies higher than 300 Hz. Table I shows the measured frequencies on the autospectrum coming from a standard measurement and the computed frequencies coming from the previous calculated model. TABLE I. Measured and simulated frequencies.

Measured frequencies (Hz) Calculated frequencies (Hz)

FIGURE 3. Definition of the roughness Ra.

It is then clear that for a given mode, the higher is the excitation, the higher will be the energy of vibration in this mode. The prototype allows us to modify precisely the measuring conditions. All the tests were provided under a sliding speed of 89 mm s⫺1 and the autospectrum were computed with 200 spectra. The challenge was then to link data coming from the different vibration modes to the characteristics of the tested surface. First, the pertinent modes have been chosen for our measurement from more than 10 visible peaks. Second, a representation of the results is proposed in order to classify clearly the different tested surfaces.

Mode 1

Mode 2

Mode 3

35 36

127 128

258 260

Moreover, Figure 4 presents the mode shapes for the first three modes. We observe that the first mode (36 Hz) moves according to an interesting horizontal oscillation on the surface of the tested fabric. This phenomenon is largely attenuated for the modes 2 (128 Hz) and (260 Hz). This means that mode 1 is probably more sensitive to friction than the other two.

FIGURE 4. Mode shapes.

DECEMBER 2005

829

Mode 2 (128 Hz) was eliminated because of the low signal/noise ratio in the autospectrum (Figure 2). SURFACE STATE REPRESENTATION When several surfaces need to be compared, the results can become difficult to analyze. To solve this problem, a two-dimensional representation of the results was proposed in which the energy of mode 3 (260 Hz) was plotted as a function of the energy of mode 1 (36 Hz). Values are standardized with relation to the maximum value. This representation is a type of map and is described in Figure 5. This map allows us to obtain a very clear analysis of the observed difference between the tested samples. TESTED

SURFACES

The tested surfaces were nonwovens meant for baby hygiene. These nonwovens were dry tested (without lotion). The technical characteristics of these nonwovens are shown in Table II. Samples Spun and Sofspan are spunbonded nonwovens (Figure 6) and samples J1 to J4 are Spunlace products (Figure 7). These last samples are designed to be impregnated with different kinds of lotions. As they are of the same structure, the samples were all produced using the same manufacturing line. Due to their large structure heterogeneity, the comparison of these different nonwovens using scanning elec-

FIGURE 5. Cross representation map.

tron microscopy (SEM) photographs was not possible. First, Spunbonded and Spunlace products were separately investigated in the present study and then, for the Spun and Sofspan products, the effects of grammage (areal density g m⫺2) variation and process modification on the surface state were observed. The Spunlace products (J1 to J4) were compared two by two, noticing the effect on the surface state, of a variation of grammage for the nonwovens J1 and J2 and of the composition of nonwovens J3 and J4 with similar grammages.

Results and Discussion To have a better understanding of the measurement data, a “time–frequency analysis” of the signal was used. This treatment was carried out by the spectral analyzer. This three-dimensional analysis allows us to observe the evolution of the spectra relative to the rubbing time. Figure 8 shows an example of the evolution of the signal for the sample J4. It can be observed that the mode 1 vibrates the whole time, whereas mode 3 vibrates only twice per cycle on the carrier (white surrounded events). When observing the blade rubbing on the surface, these events occur when the sensor is parallel to the fiber bundles on the structure. These bundles arise from the manufacturing process and are oriented along the production line. This phenomenon is visible on all the nonwovens J1 to J4. On the surface of these nonwovens the bundles are assumed to constitute a macro factor of roughness. Their presence induces periodic shocks that excite the sensor at each frequency of the spectrum. Moreover, as shown previously, mode 1 vibrates with a horizontal oscillation on the “blade–nonwoven” contact area; it is then natural to observe a continuous vibration for the whole time. However, as shocks excite the sensor at each frequency of the spectrum, macro roughness will also excite mode 1. It is thus possible to conclude from these results that the energy of mode 1 is dependent on a combination of friction and roughness, whereas the energy of mode 3 seems to be preferentially excited by the roughness of the tested nonwovens. The results (Figures 9 and 10) illustrate clearly these conclusions. In fact, when the grammage decreased for the same

TABLE II. Nonwoven characteristics. Name of the sample

Structure

SPUN 18 GSM SOFSPAN 20GSM SPUN 20 GSM J1 J2 J3 J4

Spunbonded Process 1 Spunbonded Process 2 Spunbonded Process 1 Spunlace Spunlace Spunlace Spunlace

Components 100% 100% 100% 100% 100% 65% 100%

Polypropylene Polypropylene Polypropylene Polyester Polyester Rayon, 35% Polyester Rayon

Grammage 18gm⫺2 20gm⫺2 20gm⫺2 75gm⫺2 38gm⫺2 50gm⫺2 45gm⫺2

830

TEXTILE RESEARCH JOURNAL

FIGURE 6. SEM photographs of Spun 18 GSM, Sofspan 20 GSM and Spun 20GSM.

nonwoven raw material (Spun 18 GSM and 20 GSM or Spunlace J1 and J2), only the energy of mode 3 decreased. This result is related to a decrease in surface roughness. However, when the composition of the nonwoven changes (i.e. case of the Spunlace J3 and J4), the energies of both modes 1 and 3 are modified. For example, the addition of rayon to polyester drastically increases the energies of modes 1 and 3. Such a modification of composition induces a modification of friction behavior

and also a modification of roughness [which is a combination of the hairiness properties and density and compressibility of detected macro asperities (bundles)]. In the case of the Sofspan 20 GSM products, a modification of the manufacturing process has an effect on the surface state of the tested products. This surface state depends on the calendering intensity of the process, the fiber fineness and the type of polymer used. All these factors determine roughness, friction and hairiness properties on the surface of the tested specimens.

DECEMBER 2005

831

FIGURE 7. SEM photographs of Spunlace J4. FIGURE 9. Measurement results: Spun and Sofspan nonwovens.

FIGURE 8. Time–frequency analysis.

Moreover, it is important to notice the high sensitivity of the sensor since it can detect a surface modification due to a grammage variation of 2 g m⫺2 only, in the case of Spun 18 GSM and 20 GSM products. When standardizing all the data, it is interesting to enlarge the presentation scale of our results in order to compare the Spun products with the Spunlace products (Figure 11). This representation be-

comes very useful in order to obtain a “quality map” of different products. [Note: the values of Spunbonded product move to a new quadrant in Figure 11 because they have been standardized in relation to the maximum value J4]. This enlarged representation makes it possible to differentiate the two types of nonwovens and confirms firsty that the differences between Spun products and Sofspan

832

TEXTILE RESEARCH JOURNAL face. The frequency analysis of this vibration allows us to compute the energies of each mode. Only the energies of mode 1 and 3 were studied herein. It has been shown that the energy of the first mode is sensitive to both friction and roughness whereas the energy of mode 3 is sensitive to roughness only. Several nonwovens, destined for use for hygiene products, were tested. These fibrous structures are characterized by complex surfaces. The results have clearly shown that the present test method is able to detect the effect of a small grammage variation, the effect of manufacturing processes and the effect of component modification. These modifications have been discussed in terms of “friction roughness” criteria. A modal analysis of the sensor provides a better understanding of the mechanical behavior of the plate in each frequency mode. As other work has shown that this test method is reproducible, we therefore, wish to focus on the industrial application of the present study. FIGURE 10. Measurement results on Spunlace products.

ACKNOWLEDGEMENTS Thanks to the Fiberweb and Jacob-Holm companies for the nonwovens, and to the Agence Franc¸aise de l’Innovation (OSEO-ANVAR) for the financial support.

Literature Cited

FIGURE 11. Comparison between Spunbonded and Spunlace products on an enlarged scale.

are largely lower than those observed between the Spunlace products. Second, the results show that the roughness of Spunlace products is generally higher than those of Spun and Sofspan products. Moreover, the friction of the Spunlace products is more important in comparison with the Spunbonded nonwovens.

Conclusion This study has presented a patented method capable of evaluating all kind of surface states. The developed sensor is a preloaded thin blade that vibrates according to the eigenvalues of frequencies, when rubbing on a sur-

1. Bueno, M. A, Durand, D., and Renner, M., Optical Characterization of the State of the Fabric Surfaces, Opt. Engng, 39 (6), 1697–1703 (2000). 2. Bueno, M. A, Fontaine, S., and Renner, M., “Dispositif pour e´valuer l’e´tat de surface d’un mate´riau et proce´de´ de mise en œuvre dudit dispositif”, Brevet N°4FR 0007490, International Extension in progress (PCT), 2000. 3. Bueno, M. A., Viallier, P., Durand, D, and Renner, M., Instrumental Measurement and Macroscopical Study of Sanding and Raising, Textile Res. J. 67 (11), 779 –787 (1997). 4. Poudeyimi, B., and Ramanathan, R., Measuring Fiber Orientation in Nonwovens, Part l: Simulation, Textile Res. J. 66 (11), 713–722 (1996). 5. Poudeyimi, B., and Ramanathan, R., Measuring Fiber Orientation in Nonwovens, Part II: Direct Tracking, Textile Res. J. 66 (12), 747–753 (1996). 6. Poudeyimi, B., and Sobus, J., Evaluating Carpet Appearance Loss: Surface Intensity and Roughness, Textile Res. J. 63 (9), 523–535 (1993). 7. Poudeyimi, B., Dent, R., and Davis, H., Measuring Fiber Orientation in Nonwovens, Part III: Fourier Transform, Textile Res. J. 67 (2), 143–151 (1997). 8. Poudeyimi, B., Xu, B., and Wehrle, L., Evaluating Carpet Appearance Loss : Periodicity and Tuft Placement, Textile Res. J. 64 (1), 21–32 (1994). 9. Salvado, R., “Relationship between spunbond process, structure and properties of nonwovens for hygie`ne applications”, PhD Thesis, University of Mulhouse (France) and of Beira Interior (Portugal), 2002