Optical investigations into wetted commercial coalescence filter using 3D micro-computer-tomography

https://doi.org/10.1016/j.ces.2021.117096Get rights and content

Highlights

  • 3D micro-computer-tomography imaging of dry and wetted fibrous filter.

  • Inhomogeneous local solidity responsible for inhomogeneous wetting in oleophilic filter.

  • Post processing routine developed to analyze filter, based on coarse spatially resolved images.

Abstract

Coalescence filtration is used in a variety of process industries to remove liquid mists from gas flows. Non–woven materials are widely used due to their high separation efficiency and low cost. The operating conditions of these filters are dependent on the internal structure of the filter material. Inhomogeneities, arising from the manufacturing process, not only affect the flow of air through the filter but also determine which areas are preferentially wetted, and in turn modify the airflow through the blocked regions.

This study uses three–dimensional (3D) micro–computer–tomography imaging to analyze, in a non–destructive manner, the internal structure of a commercial coalescence filter, made of oleophilic glass fibres in dry and wetted state. Using image processing software, a post–processing routine is developed to identify filter parameters based on coarse spatially–resolved images. Afterwards, the dry–filter images are compared with wetted–filter images to establish that inhomogeneous local solidity is responsible for inhomogeneous wetting in an oleophilic filter.

Introduction

Filtration represents an important step in process–engineering operations and everyday life. Besides classical solid–liquid filtration, the separation of aerosols from gas flow is also of great relevance. It comprises the purification of finely dispersed liquid droplets, such as oil droplets, from a gas that carries them. This process is particularly necessary for breathing protection, exhaust air purification, or air treatment for clean rooms. Coalescence (or mist) filtration is a widely used and cost–effective method for the separation of aerosols (Sutherland, 2008) using fibrous filter media. In addition to optimizing the most important properties of a fibre filter, namely pressure drop and separation efficiency, it is also important to reduce development and operating costs. Accordingly, the investigation and improvement of filter media properties are of great interest.

Numerical flow simulations offer, over complex experiments, the advantage of a time–saving and cost–effective investigation of the filter behavior, for a simple variation of the process parameters. While there are ample analytical models (Happel, 1959, Kuwabara, 1959, Kirsch and Fuchs, 1967, Yeh, 1972, Yeh and Liu, 1974) and empirical models (Davies, 1953, Raynor and Leith, 2000, Frising et al., 2005, Mead-Hunter et al., 2014) available in the literature for the description of a fibrous filtration process, there are only a few investigations available that compare the behavior of the filter model to real filters (Jaganathan et al., 2008). The reason for this is the lack of investigations into the internal micro structure of a real fibrous filter. Studies into filter clogging have been published by various authors before (Charvet et al., 2010, Contal et al., 2004, Gougeon, 1994, Walsh et al., 1996, Kampa et al., 2014). Some authors like Moelter and Fissan (1997) or Schweers and Löffler (1994) have studied the inhomogeneity of the fibre–size distribution in detail and its effect on the pressure drop and particle capture. All these studies use two–dimensional (2D) surface photographs from scanning electron microscope (SEM) to infer the complex three–dimensional (3D) structure inside the filter. This method clearly cannot capture any changes of the filter structure and droplet deposition in the depth of the fibrous filter, though. In addition, this technique may change the filter structure and/or the liquid distribution due to the cutting and preparation of the sample, i.e. the method is invasive. More recently, micro–computer–tomography (micro–CT) has offered the possibility for a non–invasive examination of the filter structure within the micrometer range (Schladitz et al., 2006, Lautensack et al., 2008, Hellmann et al., 2015, Lehmann et al., 2016, Tafreshi et al., 2009). In parallel, magnetic–resonance imaging (MRI) has also shown promising results (Lehmann et al., 2005, Hoferer et al., 2006). Moreover, digital volumetric imaging (DVI) has been employed by Jaganathan et al. (2008) to reconstruct 3D structures from 2D images of filter media. These authors also perform flow simulations within the 3D structures. Recently, commercial software as GeoDict© and FlowDict© became available, which have been developed over the past years to characterize and develop fibrous filters (Lehmann et al., 2016), and to simulate the flow through these filters (Gervais et al., 2015).

Independently, Chaudhuri et al., 2019, Chaudhuri et al., 2021 and Abishek et al., 2017, Abishek et al., 2018 developed tools for the random creation of (representative) 3D fibre structures that subsequently can serve as basis for computational fluid dynamic (CFD) simulations with any (commercial) code. Up to now, in the work of e.g. Chaudhuri et al., 2019, Chaudhuri et al., 2021 fibres have been modeled as straight cylinders of a single fibre diameter, and the generated filter structure at the surface closely resembles a 2D SEM image of the fibrous filter. Fig. 1 depicts such a SEM image of a sample used in Chaudhuri et al. (2021). The same filter material will be used in the present study to enable a direct comparison between results obtained from 2D and 3D imaging techniques. As is evident from the 2D SEM image, the filter has a complex and heterogeneous structure consisting of both straight and curved fibres, of varying diameters. It is obvious that only fibres within the front (focal) plane are clearly visible, while those in the depth are hardly distinguishable. Chaudhuri et al. (2021) use such SEM images to infer the distribution of fibre orientation. The in–plane angles and the out–of–plane angles of the fibres are calculated based on the Hough transformation. Additionally, the mean diameter of the fibres is approximated from the 2D images using a volumetric–averaging process. All this information is necessary for the input parameters to generate the 3D random fibre structure.

Substantial improvements can be made to the 3D fibre generation tools discussed above. Examples are (i) a better approximation of the fibre orientation, (ii) replacing uniform fibre diameters with the real (e.g. bi–modal) distribution, and (iii) accounting for the fibre curvature. However, all improvements are limited by the equipment available for analyzing the filters. Many studies use SEM images (Chaudhuri et al., 2021, Chaudhuri et al., 2021, Abishek et al., 2017, Abishek et al., 2018) to approximate depth information based on 2D images of high resolution. Others use 3D imaging devices as MRI (Hoferer et al., 2006) or micro–CT (Huang et al., 2016), along with automatic fibre recognition software to obtain true 3D information. Nevertheless, the 3D techniques often remain limited to relatively thick fibres due to the poor resolution of reasonably–priced commercial devices.

Inhomogeneity and its effect on the wetting behavior for coalescing filters, though, has been hardly investigated, to the best of the author’s knowledge. Two publications have been found, namely Penner et al. (2021) and Charvet et al. (2011). Penner et al. (2021) engage microscope images in conjunction with a differential box counting method to study the micro– and meso–scale inhomogeneities and their effect onto the liquid flow through the filter medium. Charvet et al. (2011) apply synchrotron X–ray holotomography to analyze saturated filters with regard to the liquid distribution inside the filter. Charvet et al. (2011) are limited to measure the sum of fibre and liquid volumes together, and the void space, since the absorption coefficient of both their used liquid and their fibres are similar. Charvet et al. (2011) first analyzed a dry filter. Afterwards, they place the filter in a filtration setup and expose it to an aerosol–laden air flow. Once steady–state conditions are attained, saturation S remains constant, and the saturated filter is analyzed again. Saturation is defined based on the volumes occupied by liquid and pores, i.e. S=Vliq/Vpor. This method, on the one hand, does not allow for a direct comparison between the dry and the wet filter, since the filter has been moved. On the other hand, it allows for the analysis of a filter (partially) saturated in a realistic filtration setup.

In the present study, in a first part, we apply a manual routine, after the 3D tomographic reconstruction. This routine involves image processing and subsequent recognition of the fibres, based on coarse micro–CT images with a resolution in the range of the fibre diameter. This routine is used to analyze the filter for important properties such as fibre orientation using open–source software. The aim is to generate relevant data to correct and complement the existing model parameters for further simulation purposes without using automatic fibre recognition tools. These automatic recognition tools appear to be not suitable for the coarse resolution of our micro–CT images.

Since micro–CT scanners provide 3D information without disturbing the material, in a second part of this study, images of the dry filter and the same filter sample wetted with oil (without disturbing the filter), can therefore be taken to analyze the wetting behavior within the filter without any airflow. This allows for the recognition and analysis of all the phases, namely the liquid, fibre and void spaces. The setup also allows to analyze the complete saturation spectrum from 0 to 100% saturation unlike in Charvet et al. (2011), where only the steady–state saturation (mean S33%) condition could be analyzed. Until now, for macro–scale simulations, it is assumed that for oleophilic fibres there is a uniform wetting of the fibres across the filter layer (Chaudhuri et al., 2021). However, initial investigations with back–light imaging, done to explain discrepancies between numerical simulation results and experimental findings in the work of Chaudhuri et al. (2021), show that the oil wets the fibrous filter in a non–uniform manner. This phenomena has also been suggested by Charvet et al. (2011). Hence, the present article further aims to provide more detailed information on the wetting process.

Section snippets

Experimetal setup

In this chapter, the setup and the recording of the micro–CT images are explained in more detail. Firstly, the filter material used and the micro–CT system, as well as their technical data, are presented. Secondly, the setup parameters for the image acquisition and reconstruction are explained. Thirdly, a distinction is made between the two types of examinations carried out. In the first part, a dry filter is placed in a manufacturer–supplied holder and imaged for 3D examination of the fibre

Image processing and 3D reconstruction

The diameter of the fibres in the DF5 filter are in the range of a few micrometers and the spatial resolution of the micro–CT are 5.5μm/voxel for the first part, and 7.0μm/voxel for the second part of our investigations. The different resolutions occur since the different holders lead to different distances to the X–ray source. Hence, the quality of the X–ray 2D images is not sufficient to allow for a good tomographic reconstruction of the fibre structure. After tomographic reconstruction, the

Determination of fibre orientation

The orientation of a fibre can be fully captured by two angles. The in–plane angle α represents the planar orientation of the fibre within the x-y plane (filter plane), while the out–of–plane angle β describes the inclination in the z direction (out of the filter plane). A widely–used technique to analyze the orientation of lines from 2D images is by using the Hough transformation. This transformation allows for a statistical analysis of the prevailing angles. Since the fibre–recognition

Conclusion

In this work, a glass–fibre filter for the separation of oil droplets (or aerosols) from a gas stream is experimentally investigated and evaluated. For this purpose, 2D X–ray images of an empty filter section are taken with a micro–CT and a 3D tomographic reconstruction of the filter volume is performed to characterize the structural properties. Due to the small fibre diameter of some of the fibres of the DF5 fibre filter of only a few micrometers, and due to technical limitation of the

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgement

The authors would like to thank for the cooperation with Prof. Dr.–Ing. Norbert Kockmann and M.Sc. Julia Schuler from the Group of Apparatus Design, Biochemical and Chemical Engineering, TU Dortmund. Making the micro–CT scanner available for use and (initial) help with the process of scanning and image reconstruction is greatly acknowledged.

References (46)

  • P.-C. Gervais et al.

    Simulations of filter media performances from microtomography–based computational domain

    Experimental and analytical comparison, Computers and Fluids

    (2015)
  • X. Huang et al.

    Skeleton–based tracing of curved fibers from 3D X-ray microtomographic imaging

    Results in Physics

    (2016)
  • S. Jaganathan et al.

    A realistic approach for modeling permeability of fibrous media: 3-D imaging coupled with CFD simulation

    Chem. Eng. Sci.

    (2008)
  • D. Kampa et al.

    Pressure drop and liquid transport through coalescence filter media used for oil mist filtration

    Int. J. Multiph. Flow

    (2014)
  • M.J. Lehmann et al.

    MRI as a key tool for understanding and modeling the filtration kinetics of fibrous media

    Magn. Reson. Imaging

    (2005)
  • R. Mead-Hunter et al.

    Aerosol–mist coalescing filters – a review

    Sep. Purif. Technol.

    (2014)
  • B. Mullins et al.

    Capillarity in fibrous filter media: Relationship to filter properties

    Chem. Eng. Sci.

    (2007)
  • T. Penner et al.

    Characterization of mesoscale inhomogeneity in nonwovens and its relevance in the filtration of fine mists

    J. Aerosol Sci.

    (2021)
  • P. Raynor et al.

    The influence of accumulated liquid on fibrous filter performance

    J. Aerosol Sci.

    (2000)
  • K. Schladitz et al.

    Design of acoustic trim based on geometric modeling and flow simulation for non-woven

    Comput. Mater. Sci.

    (2006)
  • E. Schweers et al.

    Realistic modelling of the behaviour of fibrous filters through consideration of filter structure

    Powder Technol.

    (1994)
  • P. Soltani et al.

    Effect of 3D fiber orientation on permeability of realistic fibrous porous networks

    Powder Technol.

    (2014)
  • J. Viguié et al.

    Finding fibres and their contacts within 3D images of disordered fibrous media

    Compos. Sci. Technol.

    (2013)
  • Cited by (5)

    View full text