Characterization of needle felts for cleanable dust filtration

1  Introduction
The fine dust fraction in dust emissions is particularly important from the point of view of air hygiene. Interest is focused on the low sinking velocities of sub-micron dust particles (resulting in a relatively long residence time in the earth’s atmosphere), their respirability and their ability to combine adsorptively with harmful substances like heavy metals. Because of their outstanding dust collecting characteristics, including in the fine dust range, cleanable dust filters are being used increasingly in virtually all sectors of exhaust gas dedusting as well as for the collection of useful products in powder form from gases.
 
Operating conditions, such as temperature peaks or temperatures that drop below the dew-point, material variables, such as the composition of the gas to be cleaned (acidic gas components, water vapour content), agglomeration characteristics and reactivity of the dust, its particle shape and particle size distribution, the structural design of the filter plant (geometry of the filter elements), and the mode of operation of the cleaning system are, among others, characteristic variables that have a crucial influence on the filtration and regeneration characteristics of filtering dust collectors. As a whole, it must be stated that although cleanable dust filters have now reached a high level of development and are being used successfully for ever more demanding dust collection problems the methods used for filter material selection and for the design of cleanable dust filters still exhibit many empirical components that are based on the “trial and error” principle. In spite of this it is, in practice, possible, through extensive experience and, in some cases, laborious pilot tests, to produce cleanable dust filter plants that operate cost-effectively with low clean gas dust contents and long service lives, even in difficult cases.
 
The filter media used for dedusting purposes are mainly cleanable sheet textile materials, such as woven fabrics and, to an increased extent, felts (needle felts, hydroentangled nonwovens). Nowadays, numerous filter media have additional surface finishes that range from surface-compacting finishing methods (calendering) to special finishes (surface coatings). They generally serve to improve the filtering properties, particularly optimum detachment of the filter cake during the cleaning and minimization of dust particle deposition inside the filter medium in order to prevent blockages.
 
The continuous onward development of textile cleanable dust filter media in respect of higher temperature stability, lower initial pressure drops, less tendency to blockages, reduced penetration by fine dust and the longest possible service life is partly responsible for the sharply rising market presence of high-performance cleanable dust filters in the industrial sector. The selection of a suitable filter medium is particularly important in the design of a dedusting plant as even now this is still to a great extent carried out on the basis of empirical criteria. The characterization of cleanable filter media and possible ways of evaluating their filtering properties are still inadequate. The textile parameters provided by the filter material manufacturers (thickness of the filtering medium, weight per unit area, etc.) and the material-specific properties of the fibre materials (strength, etc.) are usually not sufficient for reliable evaluation of the filtration behaviour of the filter media.
 
The surface porosity and its reduction with increasing distance from the surface on the raw gas side have been characterized in earlier work with the aid of a model pore [1–6]. The correlation between the structural parameters and the embedded mass of residual dust as well as the clean gas dust concentration – the two filtration-specific parameters were determined experimentally in a VDI 3926 filter material test stand – shows that needle felts with different surface structures can be characterized with the aid of these porosity parameters. This article provides information about a new way of explaining the structure of filter media by using light-optical microscopy image analysis with which new structural parameters can be obtained for comprehensive characterization of the surface of cleanable dust filter media. This should facilitate the evaluation of the filtration properties and help to improve the quality monitoring of surface-treated textile cleanable dust filter media.
 
2    Porosity of the layer of filter material
close to the surface
Figure 1 is a diagrammatic representation of the porosity of the layer of fibres in a needle felt close to the surface [6]. By carrying out an image analysis examination of a spot sample area Atot of the needle felt surface it is possible to find the total number of pores Ntot. A pore depth hp,i and a pore area Ai can be determined for each individual pore. A pore depth distribution Q0(hp) and a median value of the distribution hp50.0 (Fig. 2) can be calculated with the aid of the pore depth analysis. The total pore area can be calculated from the individual pore areas in accordance with Equation 1:

Ap,tot = Ntot Ai(1)
              i=l
The surface porosity E0 is obtained from Equation 2:
 
E0 =  Ap,tot(2)
         Atot
The total of the circumferences of the pores can be calculated from Equation 3:
 
Op,tot = Ntot Oi(3)
              i=l
Hp50.0, Ap,tot and Op,tot describe the porosity of the filter layer on the raw gas side of a sample of filter material of area Atot.
 
3    Particle penetration in surface-treated filter media
Transmitted light microscopic representations of the surface structure were evaluated by image analysis in order to quantify the porosity near the raw gas surface of needle felts in relation to the clean gas dust concentration [1]. Ap,tot and Op,tot can be determined by these imaging methods. As a result it is possible to calculate an average hydraulic pore diameter dh (Equation 4) that can be used as a measure of the mass of the particles that penetrate from the raw gas side to the clean gas side.
 
                      4 Ntot Ai
        4Ap,tot          i=l
dh =           =(4)
        Op,tot        Ntot Oi
                         i=l
The measuring equipment is shown diagrammatically in Figure 3.
It consists of a light-optical micropscope (LEICA MZ8) with great depth of definition, equipped with a cold light source and a high-resolution CCD camera (SONY ­XC-003P). The filter material sample is positioned on the object stage and exposed to the observation light. As the light passes through the filter material sample the light rays are scattered diffusely in all directions at the fibres and at any embedded dust particles, and are partially absorbed. The light intensity becomes weaker with increasing optical path length. The filter media pores can be detected by the differences in brightness that occur. The influence of daylight on the quality of the image can be reduced by installing a light screen around the microscope. In the transmitted light microscope representation the filter media fibres located at the surface are dark while the pores can be identified as lighter areas.
 
Colour pictures were taken at twenty different positions (each 2390  µm x 1860  µm) on a circular filter material sample (diameter 169  mm) to obtain statistically representative average values. The coloured pictures are converted to grey-scale images with the aid of a special computer algorithm. The images are converted into binary black and white images by defining a grey-scale threshold value. The dark areas characterize the surfaces of the fibres and the white areas characterize the pore areas. The brightness threshold value is found by assessing the optical match of the coloured image with the black and white image. The brightness threshold is altered iteratively until the best possible local agreement of the extent of the fibres in the two images is achieved (Fig. 4).
The pore areas that have been detected are then approximated by elliptical areas, from which the circumference Oi and area Ai are calculated. The results for different hydraulic diameters (calculated from Equation 4) of different filter media are listed in Table 1.
 
The penetration of particles through needle felts was determined in further investigations by measuring the average clean gas dust concentration over 100 filtration cycles on a Type 2 filter material test stand as defined in VDI 3926 (Fig. 5).
 
Al2O3 (Sasol Pural NF, d50.3 = 4.1 µm) was used as the test
dust. The dust that had penetrated through the filter medium and into the clean gas was collected by a downstream absolute filter. After a continuous filter material test for 100 filtration/cleaning cycles the quantity of dust that had been collected was weighed and calculated relative to the volume of gas that had been drawn through, giving the average clean gas dust concentration.
 
The average clean gas dust concentration over 100 cycles is compared in Figure 6 with the average hydraulic pore diameter. An interesting correlation is obtained between these two variables that shows that the average hydraulic pore diameter can be used as a measure for comparative estimation of the particle penetration in different surface-treated filter media.
 
It can be seen that the trend line intersects the x-axis at a certain dh value. One explanation for the abscissa intercept is the existence of a limiting pore diameter. All pores that are smaller than this limiting pore diameter remain irreversibly blocked after a few cleaning cycles with the result that no further particle penetration can take place. In order to investigate this state of affairs in detail the pore area density distribution ΔAi/(Atot*Δdh,i) was plotted against the hydraulic diameter dh,i (Equation 5) as shown in Figure 7.
 
dh,i = 4Ai(5)
         Oi
A pore blockage diameter db can be read from Figure 7. No particle penetration occurs below this pore blockage diameter – assuming the same test conditions and the same test dust. The determination of the blockage diameter can be explained using the following iterative procedure. A minimum pore diameter
db  dh,i,min of about 12 µm is chosen initially for the starting value for the iterative loop. All pore areas ΔAi between db and dh,i,max are then totalled. This results in an effective surface porosity «eff that is open for particle penetration in the filter medium under investigation.
  
«eff =  1     dh,i,max ∆Ai(6)
           Atot      db
By incorporating the model concept that for different porous filter media «eff is linearly proportional to the measured average clean gas dust concentration c it is possible to calculate the coefficient of determination of the linear regression R2 for the assumed db value.
 
If db is then varied from dh,i,min to dh,i,max different R2 values are obtained for the correlation of the effective surface porosity with the average clean gas dust concentration. These R2 values are plotted against the assumed db (Fig. 8). If the value assumed for db is too high then only the largest pores are taken into account for describing the particle penetration and as a result no linear relationship is obtained between «eff and c. If, on the other hand, the assumed db value is too small then small pores that have no influence on the particle penetration are taken into account and so again there is a non-linear «eff/c relationship. The coefficient of determination R2 of the linear regression is small in both cases. The point on the graph with the best coefficient of determination indicates the best linear relationship «eff/c, and therefore the correct value for db.

It can be seen in Figure 9 that db = 53 µm gives a good linear regression of the «eff/c relationship. This can be summarized by stating that the particle penetration in filter media with different structures can be compared relative to one another if the effective surface porosity «eff and the slope of the linear progression tg  are known – assuming that the same test dust and the same test procedure are used.
 
4    Parameters for residual dust embedded
in surface-treated filter media
A model pore can be derived if measurements of the 2D surface structure are combined with the measurements of the pore depth distribution of the fibre layers of the filter medium close to the surface. The model pore provides a diagrammatic representation of the pore volume that is visible when viewed from the surface into the depth of the material (Fig. 10). The pore depth distribution is determined automatically with the aid of a light-optical microscope (Olympus BX 61) equipped with a motorized object stage and a colour camera (Fig. 11). Figure 12 shows a typical result of this type of pore depth determination. The measured model pores for the filter media under investigation are shown in Figure 13.
 
Starting from the model pore it is possible to use Equation 7 to calculate a pore volume equivalent H (Table 1).
 
H = E0 · hp50.0(7)
 
The mass of embedded residual dust after 100 filtration and cleaning cycles was determined with the aid of filter tests described in VDI 3926 and plotted against the pore volume equivalent H (Fig. 14). This figure makes it clear that the mass of embedded residual dust has a virtually linear dependence on the pore volume equivalent. The pore volume equivalent can, in turn, be regarded as an average cylindrical pore volume relative to the total area with which it is possible to describe the particle storage capacity of the layer close to the surface.
 
This can be summarized by stating that the correlations shown in Figure 9 and Figure 14 correspond to two straight calibration lines that were obtained by the combined application of filter material tests conforming to VDI guidelines and newly developed light-optical microscopic pore size analyses. The particle penetration and the dust storage capacity of surface-treated ­filter media can be predicted with the aid of these straight ­calibration lines.
 
5    Pore inhomogeneity of needle felts
The inhomogeneity of the pores distributed over the filter medium plays an important part in the penetration of particles from the raw gas side of the filter material to the clean gas side and also has a crucial influence on the pressure drop characteristics. For deep-bed filters, for example, the particle penetration rises with increasing inhomogeneity while the pressure drop decreases. The parameter INH, which characterizes the inhomogeneity, can be calculated from Equation 8 either as the quotient of the pressure drop (for the same filtration velocity) or as the quotient of the filtration velocity (for the same pressure drop).
 
INH = Δphom/Δpinhom or INH = vinhom/vhom(8)
 
Lajos [8] was able to demonstrate by simulation that the ratio of the filtration velocities depends on the relative standard deviation /(1-«) of the packing density and the average packing density (1-«) of the filter medium. If the continuous blockage below the surface of the filter medium with increasing number of filtration cycles is evaluated as a type of deep-bed filtration mechanism then the obvious conclusion is that the relative standard deviation /(1-«) and porosity « parameters have an influence on the development of the residual pressure drop of cleanable needle felts.
 
Pore size measurements at the surface of needle felts can be used for calculating the relative standard deviation and porosity as follows:
                                                                                       
hydraulic diameter:  dh,i = 4 ·Ai(5)
                                        Oi

hydraulic individual pore area:  Ah,i = dh,i2 ·  =4 ·Ai2 ·  (9)
                                                         4           Oi      4

hydraulic surface porosity:          Ntot Ah,i (10)
                                                i=l
                                       «h =           
                                                 Atot     

average pore area fraction     «h  =   1   · Ntot Ah,i(11)
of the individual pores:       Ntot      Ntot     i=l  Atot


standard deviation:  s =    1     ·  Ntot Ah,i –  «h 2(12)
                                      Ntot –1      i=l     Atot         Ntot
                                                                                        
relative standard deviation:  RSD = s · Ntot(13)
                                                       «h
Figure 15 shows the results of the inhomogeneity analysis of different surface-treated needle felts. The determination of the inhomogeneity can also be applied to filter media that have been exposed to dust. For example, the inhomogeneity determination can be carried out using a VDI 3926 filter material test with which the residual pressure drop, the average clean gas dust concentration and the embedded mass of residual dust after 100 cycles are determined. Determination of the inhomogeneity parameters after 100 cycles can be regarded as a type of “dust conditioning” of the filter medium that can be used for better illustration of the effect of the inhomogeneous fibre structure on the dust collection. Figure 16 shows an example of the determination of the inhomogeneity of a filter medium that has been exposed to dust.
 
6    Conclusions
A method of determination has been developed for comparative characterization of the particle penetration and dust storage capacity of surface-treated needle felts. This method is based on transmitted light and reflected light microscopy used in combination with specially developed image analysis software with which it is possible to determine the two-dimensional surface porosity and the pore depth distribution at the surface of the filter medium. Two straight calibration lines are obtained when the surface porosity measurements are combined with filter test results that conform to the VDI guidelines, such as average dust concentration and mass of residual dust. These two straight calibration lines can be used to estimate the particle penetration and the dust storage capacity of different surface-treated needle felts. The relative standard deviation of the packing density and the average packing density can be employed for describing the inhomogeneity of cleanable dust filter media in the same way as for deep-bed filtration. A method of calculation has been worked out with which these parameters can be determined using the above-mentioned microscopic methods of measurement. These methods can be carried out successfully on needle felts that both have, and have not, been exposed to dust. This makes it possible to carry out a more comprehensive examination of the blockage mechanism with the aim of developing filter media with longer cycle times, lower average clean gas dust concentrations and longer service lives.