DEVELOPMENT OF EXTREMELY HIGH RATE FILTER

Abstract

The quest for clean water has been as old as civilisation. Filtration, oldest known method of water purification removes particulate material present in the water. Increas ing urbanisation needed water filtration at a municipal scale. Slow sand filters developed in the mid-nineteenth century were used for several years as the final treatment process for the public water supplies. When it was revealed that the filtrate from the slow sand filters was not safe enough with regard to the bacterial quality, disinfection was introduced as a final treatment. Quest for higher standards of living demanded still higher per capita rate and this led to the development of the Rapid Sand filters. Although, they resulted in an economy of space, yet, they required frequent cleaning through backwash and had a higher operating and maintenance costs. Efforts continued for a much more higher filtration rate and increased cleaning intervals. In the last few decades, several developments had taken place in this direction. The counter-current mode of operation (i.e., flow taking olace from coarser towards finer media, horizontally or vertically) results in a better utilisation of filter media, lesser headloss, higher filtration rates, higher influent turbidity inputs, longer filter runs, insignificant negative head development and almost negligible mud-ball formation, etc., at comparable physical quality of the filtrate. Even filtration without any pretreatment may be possible in certain situations. Based on the principle of counter-current operation, several developments took place, such as, the upflow filters, Bi-flow (or middle outlet) filters, Dual-media and Multi-media filters. Dual and Multi-media filters proved a costly option apart from the difficulty of selecting the media to conform with the counter-current size distri bution. (in) (iv) Proper selection of filter media, it's size distribution and operating conditions lead to an overall economical design of a filtration plant. Eliassen (mi), Stanley (1955) and Ives (1961) observed that as the deposit build up, velocity through more nearly clogged upper layers of the filter increased and these layers became less effective in removal and burden of removal passed deeper and deeper into the filter. Savage (1973) emphasised that deep bed filtration lengthened filter run. Ives (1969) showed that rate of headloss in the filter is a function of the specific deposit. O'Melia (1978) stated that small suspended particles were seen to produce much greater headloss than larger particles. Hsiung (1974) investigated a method for evaluating the specific deposit and indicated that the sludge from the backwash system seemed to retain the identity of the deposited floe within the filter bed. Ives (1969, 1970) developed a general form of the filter efficiency model for deep granular filters considering the changing specific surface ( for a particular shape of media grain and cylind rical capillaries) and interstitial velocity. Different investi gators (MacKrle, Ives, Shektmen Heert zes &Lerk; and Moroudas &Eisenklan) derived different models for the calculation of filter efficiency by using different values for the exponents y, z and x (Francoise and Haute, 1985) in the general expression. The model was treated extensively by Ives( 1970, 1975, 1985). Deb( 1969) used a rigorous analytical approach to estimate filter coefficient. Francoise and Haute( 1985) used modified filter coefficient model to predict the headloss during filtration of hydraulic floes, considering the bulking factor also. 0'Melia(1985) presented and applied relationships among raw water quality, pretreatment facilities and the design of packed bed filter. He stated that the particle size, particle concentration, particle surface characteristics and solution chemistry in the raw water supply have important and predictable effect on filter design. Huang (1986) proposed a model for predicting the progression of filtrates and headloss develop ments, incorporating the effect of particle size and particle distribution of influent (v) suspension on the filter performance. Willis(1972), Spink(1973), Hutchison( 1976), Culp( 1977), Tate(1980), Kawamura( 1983) indicated that direct filtration could be a viable alternative to conventional sediment ation-filtration system. Hudsen(1959), Cleasby, et al (1962, 1980, 1986), Letterman( 1980) showed the advantages of declining rate filtration over constant rate filtration, viz., higher production of water, simple operation, easy plant modifications, no high initial and maintenance cost of a fairly complex rate control system and superior water quality in certain situations. In the present studies, an attempt has been made to evaluate the performance of horizontal filters operated under counter-current mode, and also to evolve useful design parameters. In the case of filters operated with constant rate of filtration, the effects of variations of the various operative variables (viz., inlet head, influent turbidity, initial rate of filtration, depth and size of filter media) on the various filtration para meters (such as, maximum headloss, effluent quality, time rate of maximum specific deposit, initial headloss, etc.) were studied. In the case of varying (declining) rate filtration, modellings have been attempted for the rate of filtration, effluent quality and specific deposit. A methodology was evolved to evaluate the specific deposit at the termination of filter runs. An attempt has also been made to evaluate the values of the exponents in the Ives' efficiency model. A theoretical model is also attempted to evaluate the filtrate quality at any instant during the filtration, with respect to the initial filtrate quality, for given filter specifications. (vi) Experimental runs were conducted in the laboratory on the models of horizontal filter, under the varying modes of operation and media specifications. Different combi nations of size and depth of reversely graded sand (size varying from 1mm to 0.41mm and total depth varying from 20cm to 100cm) were tried. In the filter models, different sizes of the media were separated with the help of properly supported meshes so that an intermixing of the media did not take place. The media was properly packed and the meshes were sealed on the top to ensure that water passed only through the media and, thus, channelling along the top surface of the bed (or short-circuiting) was prevented. With an objective of doing filtration, without flocculation, the influent turbidity was prepared with known concentration of bentonite clay suspensions. The influent suspension was continuously stirred to have a uniform mix of the influent and to avoid the settling of the suspensions. The influent turbidity was fed through a constant head tank to the inlets of the various laboratory filter models. The filtration rates were valve controlled. The headloss in the filter were measured through several tappings, along the filter depth, connected to manometers. The headloss, influent and effluent quality of water were continuously monitored at suitable intervals of time during the filter run. The turbidity was estimated with a Nephlometer. At the termination of filter run, specific deposit along the media depth were determined by taking out undis turbed samples of clogged bed at various depths from the inlet end of the filter. The prepared plots, depicting variations of various parameters, were analysed for conclusions and hypotheses. For the counter-current horizontal filters operated under constant rate, some important conclusions include the following: As the filtration rate increases, there is a decrease in the value of the difference of the maximum headloss and the initial headloss, which indicates that for the higher filtration rates, the scope of the maximum allowable headloss decreases which in turn means that the performance of the filter deteriorates in terms of the resulting shorter (vii) filter runs. This was depicted from the plots of the ratio of the (maximum headloss-initial headloss) to the initial headloss versus the initial filtration rate, indicating that the studied filters should preferably not be designed for a filtration rate of more than 200 to 300 L. min"1. m"2 (i.e., about 2 to 3 times the rate of conventional Rapid sand filter). Similarly, the plots of the time rate of maximum specific deposit versus the initial filtration rate, showed that beyond a filtration rate of 200 to 300 L. min . iti blxi mum m"2, there was no increase in the value of the time rate of^specific deposit, which indicated that the turbidity was passing out of the filter without getting arrested. At such a higher rate of filtration (200 - 300 L. min"1. m" ), the effluent quality was found to be well comparable to that of Rapid sand filter. The maximum operative headloss can be assessed at a particular influent turbi dity load for a specified filter (for which initial headloss is known) from the plots prepared from the ratio of the (maximum headloss - initial headloss) to the initial headloss versus the initial headloss. The average initial headloss per unit rate of filtration was seen to decrease with the increasing inlet heads, as manifested by the plot of the ratio of the initial headloss to the initial filtration rate versus the inlet head. Beyond an inlet head of 2m, the above ratio, however, tended to become a constant value, indicating that as the inlet head decreased below 2m, the sensitivity of the above ratio increased. As the initial headloss increases, there is a decrease in the value of the difference of the maximum headloss and the initial headloss, which indicates that for higher initial headloss, the scope of allowable maximum headloss decreases which in turn means that the performance of the filter deteriorates in terms of resulting shorter filter runs. This was depicted from the plots of the ratio of the (maximum headlossinitial headloss) to the initial headloss versus the initial headloss, indicating that the initial headloss should not be more than 50-75cm , beyond which value, the above ratio tended to a constant value. (viii) The average influent turbidity should preferably be less than 30 unit so that the filtration rate does not decrease appreciably, as was apparent from the plots depict ing the ratio of the average filtration rate to the initial filtration rate versus the influent turbidity. The plots of the ratio of the average filtration rate to the initial filtration rate versus the inlet head, showed that a uniform filtration rate could be had if the inlet head did not preferably exceed 2m. Similarly, from the plots of the ratio of the (maximum headloss - initial headloss) to the initial headloss versus the inlet head, it was inferred that the ratio of the (maximum headloss - initial headloss) to the initial headloss was not significantly affected by the inlet head variations, implying that even a mediocre inlet head of around 1.75m would be appropriate. As the depth of the filter media increases, there is a decrease in the value of the difference of the maximum headloss and the initial headloss, which indicates that for larger media depths, the scope of the maximum allowable headloss decreases, which in turn means that the performance of the filter deteriorates in terms of the resulting shorter filter runs. This was depicted from the plots of the ratio of the (maxi mum headloss - initial headloss) to the initial headloss versus the media depth, indicating that the media depth should preferably be not more than 60cm , beyond which value the above ratio tended to a constant value. The plots of the ratio of the (maximum headloss - initial headloss) to the initial headloss versus the ratio of the minimum dimension of the filter cross section to the maximum size of the media grains, depicted a value of the later ratio preferably of less than 150 showed a better performance of the filter. This (as depicted by these plots) was because the value of the former ratio would be increased as the initial headloss decreased by increasing the maximum size of the media grain. Similarly, the plots of the ratio of the average filtration rate to the initial filtration rate versus the ratio of the minimum dimension of the filter cross section to the maximum size (ix) of the media grain indicated that a value of the later ratio beyond 150-250 gave a uniform filtration rate. For a filter run with the specified filtration rate, initial headloss and influent turbidity, the time rate of maximum specific deposit during the filtration may be known instantaneously, for a desired prefixed maximum headloss, with the help of the plot depicting the time rate of maximum specific deposit versus the ratio of the (maximum headloss - initial headloss) to the initial headloss. In the case of declining rate horizontal filters operated with counter - current mode of operation, main conclusions included the following: The filtration rate varied exponentially with respect to time of filter run. The modelling of such a variation had been presented and was of the general form Dn(q) = a + bt]. There were three distinct stages of decline in the filtration rate with respect to time of filtration. It was observed that for such a model, the coefficient 'a' was depen dent on initial filtration rate and did not vary significantly with the time of filter run, whereas rate constant 'b' varied significantly with respect to the time of filter run and other variables. For various incremental depths of the reversely graded media, the pattern of headloss variation with respect to time of filter run was seen to be of a different nature as compared to that observed in the case of the conventional Rapid sand filters. The pattern indicated that for the coarsest media layer the headloss increased with the time of filtration, whereas for the finest media layer it decreased, and for the intermediate media layer, the headloss increased in the beginning but decreased later. This typical pattern was explained by the fact that the headloss was caused due to a combined effect of the two factors, viz., deposit and filtration rate. In the coarsest media layer, the headloss increase was substantial in comparison to succeeding media layer, as it got first opportunity to arrest the turbidity and deposit were observed (x) to be maximum in this layer. In the intermediate media layer, initially the deposit factor for headloss increase dominated,thus, resulting in a net headloss increase, but lateron due to a substantial decrease in the permeability (i.e., decrease in filtration rate), filtration rate factor for headloss decrease became predominant and hence, there was a net decrease in the headloss. In the last finest media, the effect of filtration rate factor for headloss decrease was predominant, due to the lack of depositions in this media layer and also due to a heavy build up of deposit in the preceding coarser media layers because of which the permeability and, hence, filtration rate reduced significantly, thus, aggravating the headloss decrease. It was observed that the specific deposit measured at the termination of filter run declined along the filter depth in an exponential fashion with respect to the depth of media, measured from the inlet end of the filter. It was maximum for the coarsest media and least for the finest media. The modelling of such a variation had been pres ented and was of the general form [(1/ fi dJ = a +bL]. There were three stages of declining of specific deposit along the depth of the media. The duration of filter run to maintain a filtration rate of more than 100 L. min"1. m"2 increased with the depth of the media. The modelling of such a variation had been presented and was of the general form [T =a + bL +cL ]. With certain approximation in the Ives' model of filter efficiency, it had been observed that the efficiency of the filter decreased with the increase in the time of filtration. It was also noticed that due to the small magnitude of the specific deposit in the filter, the specific surface models for particular shape of particle and cylindrical capillaries were insignificant, and only the interstitial velocity model played an impor tant role. A methodology had been presented to evaluate the values of the exponents in the Ives' filter efficiency model. The value of the exponents in the Ives' filter efficiency model, viz., x, y, and z were determined as 1.7, 1, and 1respectively. (xi) A theoretical model had been developed to predict the quality of the effluent at any instant during the filtration, with respect to the initial filtrate quality, for given filter specifications. The various plots were prepared on the basis of the model developed for evaluating the filter performance. The employment of counter-current filter with horizontal flow have shown to result in higher filtration rates (2 to 3 times of conventional Rapid sand filters (RSF)), higher influent turbidity inputs ( 2 to 3 times of conventional RSF), and longer duration of filter run ( 2 to 3 times of conventional RSF). Thus, these are expected to result in a total advantageous effect of upto 25 times more than the conventional RSF, at comparable physical quality of the filtrate. Further, the additional advantages shall include a better utilisation of the depth of the filter media, a lesser headloss,no nega tive head development, no mud-ball formation, direct filtration feasible in some cases, no skilled labour required for washing operation for small package unit, economy in space, and curtailment of high initial as well as operation and maintenance costs of the backwash system and complex rate control systems of the conventional RSF, etc. The counter-current operated filters would have a problem of backwashing through sand expansion. To counter this backwash problem, it is suggested that the countercurrent filters are designed in small sized package units, in which the media can be manually cleaned by taking it out, washing and replacing it conveniently. This will be practical because the quantity of media in each package unit would be small and the sand cleaning would be needed after long intervals of time, such as 2 - 5 times a month. One or more such package units can be taken out in a cyclic order for cleaning so that the treatment to full capacity is carried out uninterruptedly. Even big units can be cleaned in this manner provided the standbye is made 100%. Radial filters may also be one of the options of the counter-current horizontal package filters, to have a compact design, thus, resulting in an economy in space. (xii) For small water supply schemes, where bulk quantity of water is not required, such as, rural water supply schemes (especially in the hills), small industrial colonies, campuses, schools and other small institutions, small package filter with counter-current horizontal flow may prove to be of immense use. Even declining rate filters may be used in place of constant rate filter to utilise high yield of the former and to avoid the complex rate control systems of the later. A number of such filters (except the standby units) can be run simultaneously in series in a cyclic order such that when the filtration rate of the first filter is at peak, that of the last one in the series may be at a predecided minimal, and the total filtrate volume from all the filters is the same as required. The channels, etc. can be designed in such a way that these can incorporate both peak and minimum flows. Backwashed units can also be replaced in the cyclic order. Such a system is expected to lead to an overall economy and development in the filtration process, and consequently in the treatment of water.