Recent Improvements in the Measurement and Control of Ceramic Foam Filter Quality B Milligan*, S F Ray** *Pyrotek SA, Sierre, Switzerland **Pyrotek Engineering Materials Ltd, Netherton, UK. www.pyrotekeurope.com Abstract The use of ceramic foam filters to treat molten aluminium is now commonplace. Increasing demands on metal quality and improvements in the application of filters have secured their use. They now offer a simple, reliable and cost effective method to remove inclusions from liquid aluminium which is an important part of meeting the quality demands of specific product applications. To avoid process difficulties in the casthouse, it is important that the filters are consistent. This need for high quality, consistent filters becomes greater, when using Grade 40 filters or finer. Filter Grades (or ppi) are normally specified using air permeability. Air permeability alone does not provide an adequate understanding of filter geometry and thus performance. Filters that have the same air permeability can have very different pore sizes and thus their filtration efficiencies vary. This paper will explain why it is important to understand and specify the pore size of filters. It will then detail the changes made to the production facilities at Pyrotek SA, Switzerland. These changes have improved the consistency of the SIVEX ceramic foam filters and also allow Pyrotek to specify filters by pore size. 1 INTRODUCTION The use of ceramic foam filters in the aluminium casthouse has led to an improvement in metal quality for many aluminium producers. Ceramic foam filters are now specified for most casting operations and the availability of consistent fine pore ceramic foam filters offers a practical alternative to deep bed systems from both a cost and quality perspective. There are several manufacturers of ceramic foam filters worldwide, and the quality of filters produced by all these has gradually increased over the last 20 years. The current situation is that all the manufacturers are able to produce a fine quality ceramic material in the replica of the starting foam. However, in the author’s opinions; too little attention has been paid to the relationship between the filter cell or pore size and the filtration efficiency of the filters. Indeed, only since the widespread utilisation of the LiMCA II system has it been possible to effectively measure the filtration efficiency, so that filtration mechanisms and in particular the effect of filter pore size can be further understood. This paper will detail some recent studies of filtration efficiency undertaken jointly by Pyrotek Engineering Materials and VAW R&D in Bonn[1, 2]. It will then show how this information has been used to both improve the understanding of the ceramic foam filtration mechanisms and thus improve the quality and consistency of the filters. 1.1 Ceramic Foam Filter Geometry Various terminologies are used to describe the size of the pores in a ceramic foam. In this paper, the terms cell size and pore size are taken to mean the same and are used interchangeably. Further explanations as to the physical properties these terms relate to are given on the following page. Ceramic foam filters are produced by impregnating a reticulated polyurethane foam with a ceramic slip, removing the excess slip by squeezing the foam, and then drying and firing the body. The result is a ceramic replica of the original foam. The structure can be regarded as a series of interconnected dodecahedra. Fig 1 shows a model of a foam structure. Fig 1a shows a single dodecahedra representing a foam cell or pore. Fig 1b shows four such dodecahedra interconnected such as is the case in a small section of a ceramic foam filter. Figure 1a - A dodecahedra representing a single cell or pore in a ceramic foam filter Figure 1b – Four interconnected dodecahedra showing a typical structure of a ceramic foam filter Figure 1 – Representations of a foam geometry The cell or pore size is the diameter of the individual dodecahedra. The window size is the diameter of the interconnecting area between the dodecahedra. The window size and cell or pore size are, as one might expect – geometrically related. This will be shown in more detail later in the paper (Figure 9). Another visualisation of foam geometry is shown in Figure 2. Here the pores are represented by spheres and these are interconnected. The pore or cell diameter is indicated by p and the window diameter by ? . Figure 2 - A schematic representation of the pores in a ceramic foam filter showing pore diameter p and window diameter ? Finally Figure 3 shows the difference between windows and pores by highlighting them on a piece of foam. For each image, a sample of polyurethane foam, treated with blue colorant was photographed. The pores and windows were then highlighted using the image analysis system described in the Experimental Procedure. Fig 3a shows the windows. Figure 3b shows the pores. Figure 3a - Windows Figure 3b - Pores Figure 3 - Showing the difference between pores and windows by highlighting them on a piece of polyurethane foam 2 EXPERIMENTAL PROCEDURE A measurement technique was developed to measure pore size. This is described here, although the system is proprietary to Pyrotek Engineering Materials Ltd and Pyrotek SA, so some details have been omitted. The measurements were made using a high performance Image Analysis System. The images were taken using a Stereoscopic zoom microscope with a magnification range from 3 to 60 times. The microscope was fitted with a 3-CCD camera linked to the image analysis system. Foam samples were treated before processing to increase the contrast of the otherwise black on black image. It was not necessary to treat filter samples. An advantage of the test is that it is non-destructive. Neither foam nor filter is damaged during testing. At least five images of each filter were taken for pore size measurement and at least five for window size. The images were taken in each corner and the centre of the piece. The lighting of the sample is very important. Different lighting techniques are required for pore and window size measurements. For pore size measurements, one must try to obtain an image as close as possible to a 2D representation of the filter surface. For window size measurements, it is necessary to light the sample so that the internal structure can be seen. 2.1 Calibration The system was calibrated by measuring a ‘graticule’ and establishing a microns per pixel scale for each magnification. These are given in Table I. Magnification Microns per pixel 3 41.3 6 20.7 8 15.5 10 12.4 12 10.3 16 7.75 20 6.2 Table I - Calibration scale for each magnification used 2.2 Measurement of Window Size Figure 4a shows typical images of foam and Figure 4b for filter lit for window size measurement. The direct lighting from above penetrates deep into the filter so the internal structure of the filter can be clearly seen. These images are ideal for measuring window size. Figure 5 shows the steps in the analysis process. 2.3 Measurement of Pore Size Figure 6 shows examples of foam and filter images lit for pore size measurement. The lighting method highlights only the surface of the filter. Figure 7 shows the steps in the analysis process. 2.4 Output of Results The results are produced in both graphical and database form by image analysis software. This means that the information can be quickly checked while working and then the database transferred to other software packages for more detailed analysis and storage. Figure 4 - Typical foam and filter images lit for window size measurement Figure 4a - Foam An example foam image suitable for window size measurement. Figure 4b - Filter An example filter image suitable for window size measurement. Figure 5 - The steps in the window size measurement procedure Step 1 - The initial image (Figure 4a) is processed to remove shading variations Step 3 - The image is thresholded and the resulting shapes superimposed on the starting image for measurement. Step 2 - The image is processed further to highlight the location of windows. The starting image is the foam image from Figure 4a. This image is smoothed to remove lighting defects (Step 1). Then the image is processed. The comp uter takes each pixel of the image and evaluates if the areas that surround it are lighter or darker than it. It then attempts to isolate the dark areas (Step 2). The darker areas are separated out of the image by ‘thresholding’ it. They are then superimposed on the starting image and measured (Step 3). Figure 6 - Foam and filter images lit for pore size measurement Figure 6a - Foam An example foam image suitable for pore size measurement. Figure 6b - Filter An example filter image suitable for pore size measurement. Figure 7 - The steps in the pore size measurement procedure Step 1 The initial image (Figure 6a) is processed to remove shading variations Step 2 The image is processed further to highlight the location of pores . The procedure here is very similar to the measurement of window size. The main difference is that the image processing operation is performed for longer. This ensures that the larger cells are clearly visible and easy to threshold to produce the mask image for measurement. Step 3 The image is thresholded and the pore outlines are super imposed on the starting image for screening and measurement 3 RESULTS Figure 8 shows the correlation between both foam pore (or cell) and window size and foam permeability. 5 cell size window size diameter (microns) Thousands 4 3 2 1 0 0 20 40 60 80 100 120 foam permeability Figure 8 - Foam cell and window size versus foam permeability Figure 9 shows the correlation between foam cell (or pore) size and window size. Points are plotted for the entire range of foam from Grades 10 to 80. A regression analysis on this correlation gives r2 as 0.993 and the gradient of the line as 2.757. 5 Thousands cell size (microns) 4 3 2 1 0 0 500 1000 1500 2000 window size (microns) Figure 9 - Correlation between foam pore size and foam window size Figure 10 shows the correlation of filter pore size and foam pore size. The regression analysis gives r2 of 0.998 and the gradient is 0.932. This indicates that the filter pores (or cells) are slightly smaller than the foam pores (or cells). 5 Thousands filter cell size (microns) 4 3 2 1 0 0 1 2 Thousands 3 4 5 foam cell size (microns) Figure 10 – Correlation between filter pore size and foam pore size 4 DISCUSSION A system to automatically measure the pore size of ceramic foam filters and reticulated polyurethane foams has been developed, and proven to be reliable. A correlation between foam pore size and foam permeability has been noted, for a given density of foam. There are also good correlations between: ?? foam pore size and foam window size ?? filter pore size and foam pore size The former confirms that there is a geometric relationship between pore size and window size. The latter shows that by carefully controlling the starting foam it is possible to control the final filter pore size. 4.1 The importance of pore size Pore size is important because it is the most critical factor controlling the priming and efficiency of a ceramic foam filter. For a filter to prime, liquid aluminium must displace the air in the pores at the surface of the filter. The pressure required to achieve this is inversely proportional to the size of the pore. Figure 11 shows the effect of pore size on the height of metal required to prime a filter. Clearly, the smaller the pore size, the greater the metal height (or metallostatic pressure) required to initiate metal flow. The graph shows a minimum and maximum priming height to compensate for various production conditions and alloy types. 450 priming height (mm) 400 350 min priming height 300 max priming height 250 200 150 100 50 0 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 pore size (microns) Figure 11 – Filter pore size versus priming height Since the metal flows through the pores, the quantity and size of inclusions captured and thus the efficiency of the filter is related to the size of the pores. Smaller pores capture smaller inclusions. Figure 12 shows a summary of some of the work carried out jointly by Pyrotek and VAW R&D[2] . The graph refers to LiMCA II removal efficiency data for 17” filters at 10 tonnes per hour in 1050 alloy with no grain refiner or degassing upstream of the filter. It shows that there is a significant difference in average removal efficiency for the two different pore size filters. Clearly the probability of capturing smaller inclusions is higher with a smaller pore size. 100 % removal efficiency 90 80 70 Filter Av. Pore Size:- 1945 microns 60 Filter Av. Pore Size:- 960 microns 50 15-20 20-25 25-30 30-35 35-40 40-45 45-50 50-55 55-60 Inclusions Sizes (microns) Figure 12 – Average % inclusion removal efficiencies for a range of inclusion sizes, and for filters with two different average pore sizes Finally one must consider that there is a range of pore sizes within any filter. The variation of pore sizes can also have a tremendous effect on the consistency of filtration efficiencies. Figure 13 is again taken from work undertaken jointly by Pyrotek and VAW R&D. In this graph an attempt to highlight the reliability of various filter grades has been made. During many trials, a range of inclusion removal efficiencies were recorded. This graph shows the maximum efficiency recorded plotted against the smallest pore size in the filter, and the minimum efficiency recorded is plotted against the largest pore size. 100 90 Filtration Efficiency (%) 80 65 50 30 80 70 60 Max Filter Pore Size vs Min Efficiency Min Filter Pore Size vs Max Efficiency 50 0 500 1000 1500 2000 2500 Filter Pore Size (microns) Figure 13 – Showing the relative reliability of filters with various pore sizes Figure 13 illustrates that whilst all filters are capable of producing good filtration performance, one must take into consideration the performance over many casts to gain a good understanding of their behaviour. Clearly, smaller pores give a more reliable filtration performance. 4.2 Relationship between pore size and air permeability Air permeability is a test method that determines the resistance to flow (or pressure drop) across a piece of reticulated foam or ceramic foam filters. It is often used to indicate the differences between filter grades. Logically, the pore size of the filter should be related to air permeability, however there are a number of other factors that work in combination with the pore size to determine the permeability of a filter. These include the amount of ceramic material used to coat the foam, the quality of the coating and the number of blocked pores. So it is quite possible to have two filters with the same pore sizes and very different permeabilities, or the same permeability and different pore sizes. It is therefore important to be able to measure and control both the air permeability and pore size of ceramic foam filters. It will be shown later that the pore size, and not the air permeability of the filter determines filtration efficiency; however, the air permeability influences the running head of the filter and therefore must be controlled within acceptable limits. Figure 14 shows this affect graphically. The three factors considered here are filter pore size, filter permeability and filter density (an indication of the amount of ceramic used to coat the foam). The graph clearly shows that it is important to monitor and control all three of these parameters. Furthermore, it is possible for filters of the same permeability to have very different pore sizes. Permeability (Pa) 700 350 0.5 0 0.46 750 1250 1750 Pore Size (microns) Density (g/cc) 0.42 2250 Figure 14 - Showing the relationship between filter density, pore size and permeability 4.3 Relationship between pore size and ppi Traditionally ceramic foam filter manufacturers have used ppi or pores per inch to indicate the pore size of ceramic foam filters. This has served well as a guide, however; since there are no standard test methods for measuring ppi, it has led to large discrepancies in pore size between various filter manufacturers. This lack of an absolute measurement of pore size has made it difficult for manufacturers to control filter quality, leading to inconsistencies in practical application. As an example of the sort of discrepancies that can be seen from commercially available ceramic foam filters, Table II overleaf shows pores size, permeability and density of two filter grades or ppi’s for the four largest filter manufacturers. Table II clearly shows the need for more reliable and consistent control of pore size and a means that can be used to ensure consistent performance regardless of manufacturer Filter Grade/ppi 30 50 Perm (Pa) Pore (microns) Density (g/cc) Perm (Pa) Pore (microns) Density (g/cc) Manufacturer A Manufacturer B Manufacturer C Manufacturer D 89 113 122 74 2453 1940 1875 2237 0.40 0.44 0.43 0.49 136 275 258 115 1320 984 1022 1367 0.41 0.44 0.44 0.49 Table II – Comparison of Filter Permeability and Pore Size for Various Filter Manufacturers 4.4 Improvements in the Filter Manufacturing Process The measurement technique developed is now an integral part of Pyrotek’s quality control system for the production of SIVEX ceramic foam filters. The system is used to check incoming foam pore size. This has allowed a revision of the specification of ceramic foam filters. There have been four changes. ?? Filters are now referred to by ‘grade’ and not ppi ?? The mean pore sizes follow a logical pattern. A grade 20 has a pore size of approximately half that of a grade 10. A grade 40 is half of a grade 20 etc. ?? The tolerances for pore sizes with each grade have been reduced leading to improved consistency ?? There are distinct gaps between filter grades These improvements are shown graphically in Figure 15 and the new specifications are shown in Table III. 10 ppi Grade 10 20 ppi Grade 20 30 ppi Grade 30 40 ppi Grade 40 50 ppi Grade 50 60 ppi Grade 65 old specification new specifications Grade 80 0 1000 2000 3000 4000 5000 6000 Average Cell Size (microns) Figure 15 - Comparison of old and new filter pore size specifications Filter Grade Average Cell Size (microns) Min Max 10 3800 5100 20 2300 2900 30 1705 2105 40 1250 1510 50 900 1120 65 710 860 80 600 700 Table III – The new specifications for ceramic foam filters 5 CONCLUSIONS A technique for measurement of ceramic foam filters and reticulated polyurethane foams has been developed, and proven to be reliable. ?? Technique successfully introduced as a Quality Control tool ?? New filter specifications have been developed to improve filter quality and ultimately metal quality The influence of pore size on filter performance has been presented, notably: ?? The relationship between pore size and priming height ?? The effect of pore size upon the probability of inclusion capture, by size of inclusion ?? The reliability of filters as a function of pore size The interaction between pore size, air permeability and filter density has been discussed. ?? It is possible to have the same permeability, but very different pore sizes ?? It is therefore important to control all these parameters during filter production. ?? Only by controlling all these parameters can the reliability of ceramic foam filters be improved 6 REFERENCES [1] N J Keegan, W Schneider, H P Krug, ‘Evaluation of the Efficiency of Fine Pore Ceramic Foam’, Light Metals 1999, pp 1031 – 1041 [2] N J Keegan, W Schneider, H P Krug, ‘Efficiency and Performance of Industrial Filtration Systems’, 6th Australasian Asian Pacific Course and Conference, Aluminium Casthouse Technology: Theory and Practice (ED: M Nilmani), TMS 1999, pp 159-174 [3] S Ray, N J Keegan, ‘Measurement of Cell and Window Size in Ceramic Foam Filter Manufacture’, Light Metals 1998, pp 885 - 894