Methods for Paper Structure Characterisation by Means of Image Analysis

Abstract

Paper is a complex, disordered porous media. The complexity makes it difficult to obtain more than empirical knowledge of the relationship between papermaking variables and the resulting paper structural properties. It is known that the alteration of the paper structure is decisive for many paper properties, but the appropriate tools for visualising and assessing the changes to the paper structure after varying physical treatments of the paper have until recently not been available. Scanning electron microscope in backscatter mode (SEM-BEI) and X-ray microtomography allows acquisition of respectively cross-sectional and 3D images of the detailed paper structure with sufficient contrast and resolution to clearly discern the structural constituents. The access to detailed images of the paper structure allows quantification of the changes made to the detailed paper structure through application of suitable image analysis routines. The microscopy techniques combined with image analysis thus provide the tools needed to improve the knowledge of how various papermaking variables may improve the paper properties through alteration of the paper structure. Improvement of the knowledge of the structural mechanisms governing the paper properties may help to give a more knowledge based operation of the paper machines and may be applied for process and product optimization.

This thesis work has considered different aspects of how the physics governing the SEM-BEI technique, image acquisition and sample preparation may affect the quality of the cross-sectional images of the paper structure with regard to the subsequent image analysis. The effect of the applied filtering and segmentation routines to obtain the binary representation needed for image analysis are also considered.

It was found that the applied methodology yielded realistic binary cross-sectional representations of the paper structure and thus representative structural assessments. However, to obtain assessments that are representative for the paper grade and not only the small cross section, at least 15 independent cross-sectional images were needed. Then umber of necessary replicates varied somewhat depending on the structural characteristics.

This study involved improvement of existing image analysis techniques. The main contribution from this work in this aspect is the application of the rolling ball algorithm for an objective surface definition of the physically small and often uneven paper cross sections. The objective surface definition enables the division of the structure into locally uniformly thick layers, for which the material distribution in the z-direction of the paper can be determined. The material distributions that may be determined include solid/void fractions, fillers and fines content. The improved image analysis routines also include measurement of local thickness, density and basis weight, pore chords and specific surface area.

The practical application of the SEM technique for obtaining new knowledge of papermaking variables: paper structure: paper properties relationships included assessment of the pore geometry and the material distribution in the z-direction. The z-directional material distribution was determined in order to assess the effect of temperature gradient calendering on a SC-paper grade. The results obtained for the z-directional material distributions combined with surface topography techniques revealed that the temperature gradient effect was concentrated to the very few outermost micrometers of the paper structure.

The pore geometry was assessed by both the 2D and 3D techniques. The determined pore chord distributions followed the theoretically derived pore height distribution of Dodson for the z-direction. The pore chord distribution for any spatial direction appeared to follow a log-normal distribution, when including all pore chord sizes to allow a “continuous” distribution. The mean and standard deviation were proven to be proportional, as earlier observed from physical pore size measurements and theoretical deductions. The width of the pore chord distributions for the different spatial directions changed with the structural anisotropy. The wider the pore height distribution, the more different are the pore chord mean and standard deviation.

Most of the information obtained for the pore geometry from assessment of pore chords can be summarized in the equivalent pore representation. The ellipsoid shape of the equivalent pore, constructed as a warped surface of the mean pore chords for all spatial directions, was confirmed from the assessments of 3D images of three different paper grades having distinctly different structural properties. The ellipsoid shape was also found for the pore chord standard deviation. Additionally, it was found that the ellipsoid constructed for the solid phase had the same shape as for the porous phase and their relative size was proportional to the solid fraction of the assessed paper structure. The equivalent pore ellipsoid thus contain all information about the pore/fibre chord distributions in all spatial directions in addition to the structural anisotropy.

The studies of the pore geometry prove that the solid fraction and the structural anisotropy are the major determinants for the pore geometry.

The results reveal that there is clear differences between the pore size distribution determined by image analysis and physical assessments like mercury porosimetry, which is affected by narrow pore necks limiting the intrusion from the surfaces.

This thesis work have applied three different applications of X-ray microtomography. The three different applications include both phase contrast and absorption mode contrasting for high resolution synchrotron source X-ray microtomography and low resolution X-ray microtomography provided by a stationary source commercial scanner. The experiences from this work have proven that the synchrotron source X-ray microtomography can be applied successfully to obtain high quality digital representations of the 3D paper structure with sufficient contrast and resolution to detect and quantify the details of the fibre network. The current technology at the applied beam lines at the ESRF, France, allows spatial resolution down to approximately 1 µm. This resolution is sufficient for obtaining realistic structural assessments that are suited for determining how the paper structure is affected by papermaking variables and itself influences the paper properties. The high resolution also provides insight into the detailed paper characteristics from visualization of the 3D paper structure.

The resolution provided by the stationary X-ray source is not sufficient to preserve the fibre network topology. However, the signal is of sufficient quality to make comparative structural assessments possible, although the numerical results from the structural assessments are not physically reasonable.

This thesis work have considered different aspects of how the working principles, image acquisition, sample preparation, sample mounting and volume reconstruction of the three applied X-ray microtomography techniques may affect the quality of the 3D images of the paper structure with regard to the subsequent image analysis. It was found that the sample mounting for the high resolution absorption mode images was clearly affecting the digital representation of the paper structure. An alternative sample mounting, avoiding the melt glue penetrating into the paper structure, is recommended for future application of X-ray microtomography. However, it must be emphasized that restricting sample movement during the image acquisition is of outmost importance to obtain 3Dimages with a minimal content of noise.

The applied routines for image filtering and segmentation to obtain binary representations are presented and the quality of the resulting binary structures are evaluated. The results prove that the applied image processing preserves most of the information provided by the greyscale images. However, the applied routines are not optimised. Exploiting more of the three-dimensional connectivity in the images and applying edge preserving filtering techniques may improve the image processing for future experiments. The image processing of the phase contrast images is difficult, semi-automatic, laborious and computationally demanding due to the relatively low contrast and the contrasting technique detects only the phase borders. The image processing of the absorption mode images is considerably more straightforward. The quality of the binary absorption mode images is almost as good as the binary phase contrast images, despite the lower resolution. The absorption mode is therefore recommended for the future application of high resolution X-ray microtomography. Additional benefits of the absorption mode imaging is the ability to discern mineral particles and coating and that it has no principal problems in acquiring images of high density paper grades. A benefit only provided by the phase contrast technique is that it provides sufficient contrast between wood fibres and water to acquire 3D images of soaked paper samples.

This study has applied and looked into the details of many image analysis routines and transport simulations to determine the characteristics of the imaged paper structures. Most of these routines have already been applied to low resolution 3D images of paper or to 3D images of geological samples. The main contribution from this study regarding the analysis of the structure is thus modification of some of the image analysis routines to include the surface layers after defining the surface using the rolling ball algorithm.

The practical application of the X-ray microtomography technique showed that calendering of a newsprint-like paper grade had a dominating effect on the paper structure compared to addition of reinforcement pulp, addition of retention aid and alteration of the head box consistency. The calendering resulted in a denser sheet having smaller pores and a higher resistance against transport through the porous phase, especially for the z-directional flow. The analysis of the applied factorial design yielded indicative results for the other papermaking variables. The few significant effects revealed that the addition of reinforcement fibres yielded slightly larger average pores and the addition of retention aid resulted in a slightly more complex structure. The alteration of the head box consistency seemed to have an insignificant effect on the small paper volumes.

Additionally, the assessments showed that the standard deviation for the assessed structural characteristics were less than ±10% of the mean value. The mean values for the determined transport properties showed a higher standard deviation. However, the size of the 3D digital paper volumes render division of the volumes into sufficiently large sub volumes to maintain representative scales possible. Assessment of the transport properties of a number of sub volumes enabled a determination of the transport property: porosity relationship instead of a single mean value. In this way more information was extracted from the volume and the results were less affected by normal variation. The results from the practical study of X-ray microtomography confirm the hypothesis that the density/solid fraction of the paper structure is a major determinant of the paper properties.

An important part of this study has been to consider which is the right technique of the2D and 3D methods for various structural characterization applications. There is a clear differentiation between the practical applications of the techniques. The SEM technique is the preferred technique for more repetitive and routine type assessments, for characterizations where the 3D extension of the objects are not of crucial interest or for assessments where the nature of fines and fibrils are of high importance. The X-ray microtomography is a natural choice for more complex structural assessments where the extension of the fibres and pores are of high interest, as for e.g. assessment of transport properties. The 3D extension of the structure may also be needed to obtain a better knowledge of how the paper structure influences the bulk paper performance properties than provided by the cross-sectional approach. X-ray microtomography is thus a powerful research tool that can provide information which is difficult or impossible to access applying other methods. However, the technique has its clear limitations when it comes to availability, representativity and resolution. It is also important keep in mind that the technique will not in the near future be an everyday technique that many scientists will have access to.

This study have helped establishing the SEM-BEI and X-ray microtomography microscopy combined with image analysis for paper structure characterisation. The tools provide many possible applications for the paper researchers that see the available possibilities. However, the results from application of the techniques for determining relationships between papermaking variables and paper structure characteristics, and possible relationships between structural characteristics and paper properties, are never better than the applied experimental scheme. Future application of the techniques for structural characterisation will hopefully add to the knowledge of the detailed properties of the paper material and thus assist in product and process optimization.