Optimization of Solid/Liquid Separation Processes Using Particle Size Distribution Analysis

A study was conducted to investigate the influence of the coagulant dosage, suspension pH, mixing intensity and time on the suspended solids particle size distribution during coagulation, flocculation and filtration of dilute suspensions. The coagulation/flocculation investigation consisted of laboratory experiments utilizing the jar test ' apparatus as well as a pilot plant mixing tank. Filtration experiments were conducted using 3-six inch inside diameter glass columns with different media sizes. The data · generated were used to compute the distribution slope of the particles in each sample. The distribution slope was used as the pretreatment performance indicator. The distribution slope is influenced by both chemical and physical treatment parameters and it can be used in the treatment of dilute suspensions to optimize the coagulant dosage, mixing intensity, and mixing time. Both the total number of particles and the distribution slope of flocculated suspensions were found to follow a second order polynomial equation due to floe breakage after the optimum mixing time. The coagulant dosage was found to affect both the head loss development and the effluent quality of granular media filters. The mixing intensity and time were found to influence mostly the head loss development and not the effluent quality. The effluent quality improved with the increase in the coagulant dosage and mixing time and deteriorated with the increase in the m~xing intensity. The filter's effluent particle size distribution was influenced by both the influent particle size distribution ' and media size. Fine media was sensitive to the influent changes whereas coarse media showed much less sensitivity to changes in the influent characteristics.

The filter's effluent particle size distribution was influenced by both the influent particle size distribution ' and media size. Fine media was sensitive to the influent changes whereas coarse media showed much less sensitivity to changes in the influent characteristics.                    110 -Single colle-ctor efficiency due to diffusion 111 -Single collector efficiency due to interception lls -Single collector efficiency due to sedimentation  (1) contact filtration, in which destabilizing chemicals are added to the raw water and the resulting destabilized suspension is applied to the filter without further processing; ( 2 ) direct filtration, in which suspension is flocculated to growth prior to filtration; and the destabilized accomplish particle (3) conventional treatment, in which destabilization, fl~cculation, and sedimentation precede filtration.
Contact filtration is optimal at very low particle concentrations (less than 5-10 mg/L). This is because excessively long flocculation periods are needed to produce particle aggregation in such dilute suspensions. Direct filtration is economically attractive when the mass concentration of particles in the raw water supply is less than 15 to 20 mg/L. Flocculation in this process provides sufficient particle growth to reduce the head loss and breakthrough caused by small particles. Conventional treatment is usually practiced when the mass concentrations of the particles in the raw water is greater than 20 mg/L. Silverman et al. (1983) found that turbidity is closely associated with particle sizes around 2-40 micrometers.
Particle counts, in contrast to turbidity measurements, provide a direct measure of the particulate matter present in the water and its size distribution. The particle size distribution can be used to record the growth of aggregates in the flocculation step and to predict their subsequent removal by direct filtration. Particle counting is a more sensitive analytical technique than turbidity for measuring the concentration of suspended solids greater than 1 micrometer in particle diameter or for raw water with low turbidity values (less than 5 turbidity units) Monscvitz et al. 1983).
According to Tate et al.(1978) and Hutchinson (1985), particle counting has the following advantages over turbidity measurements: (1) a direct measurement of particulate material is provided, (2) the sensitivity of suspended solids measurements is enhanced, The results obtained are limited in application but they can be useful in the elimination of relatively ineffective variables for subsequent pilot-plant testings.
The use of particle size analysis has been limited due to the cost associated with analytical equipment and the time required for data analysis. Moreover the use of particle size distribution analysis in water treatment plants has not received great attention due to the lack of information about its importance. Despite these advantages, the use of on-line particle counting in treatment plants has been shown to be cost effective. Hutchinson (1985) reported as much as 30% reduction in the chemical costs was achieved when particle counting was incorporated in the 390 million gallon per day direct filtration plant supplying the Las Vegas area. Because particl~ counters have the capability of measuring different particle sizes in a sample, (up to 12 sizes), it is common practice to use just the total particle count as an indication of filter performance. Recent research in direct filtration has focused on the effect of influent properties of the suspension on the removal of such 10 parameters as turbidity, color and particle count. The effect of the influent particle size distribution on the head loss, effluent quality, and effluent particle size distribution has not been studied.

Objectives of the Study
The research reported in this study focuses on the physical aspects of coagulation, flocculation and filtration, specifically the changes in particle size distribution brought about by these processes.
Specifically, the objectives of this research are: (1) to investigate the influence of physical and chemical parameters in the processes on the dilute suspensions.
coagulation and flocculation particle size distribution of (2) to investigate the influence of influent particle size distribution on granular media filtration in terms of head loss development across the media and effluent quality (particle count), and (3) to investigate the relationship between the influent and effluent particle size distribution with different filtration media sizes. 11

LITERATURE REVIEW
In this chapter, the three most important processes in water treatment will be reviewed. The three processes are the ones used in treating dilute suspensions, mainly those used in direct filtration. These processes include coagulation, flocculation and filtration. Mixing intensity is described in terms of the velocity gradient, G, which is derived from the mean amount of work applied per unit of time to a unit volume of fluid at a definite viscosity (Camp 1955).
The third process is filtration which consists of using single media (usually graded sand), dual-media (sand and anthracite) or multi-media (sand, anthracite, and garnet sand) as filtering materials.
In conventional treatment, filtration is considered a polishing step but in direct ' filtration it is considered the most important operation unit of the entire treatment process.
The following sections describe the three processes in terms of the mechanisms involved and the traditional methods of optimization of these processes. The last section is a review of particle counting and the use of particle size distribution in the treatment of waters.

Coaqulation
Coagulation is defined as the chemical destabilization that leads to a reduction of the potential energy of (1) an imperfection in a crystal lattice (2) ionization of molecules at the particle surface

Mechanisms of Coagulation
The coagulation of particles in natural waters involves two distinct steps (Stumm et al. 1968and Dempsey et al. 1984b): (1) Particle transport to bring about interparticle contacts, and (2) particle destabilization to permit attachment when contact occurs.
Theories of particle transport are based upon fluid and particle mechanics; theories of particle destabilization are based on colloid and interfacial chemistry.
Particle transport in aqueous systems is a physical process. There are at least four different mechanisms at which particles present in natural waters can be destabilized.

Flocculation
Flocculation is the physical process of bringing the coagulated particles into contact to promote floe formation.
Flocculation of particles has been proposed to occur due to three mechanisms (O'Melia 1972, Amirtharajah et al. 1986 and Lawler et al 1983): (1) Brownian or perikinetic flocculation due to the molecular motion of water against the particle surface; (2) velocity gradient or orthokinetic flocculation due to bulk fluid motion; and (3) differential settling due to a larger overtaking and colliding with a slower particle. The relative importance of the transport mechanisms in flocculation is based on the size of particles present in suspension.
Perikinetic flocculation predominates when particles present are less than 0.1 micrometer in size (Ives 1978b)and with larger particles, orthokinetic transport predominates. Flocculation by differential sedimentation can occur in flocculators even though particle sedimentation is prevented by high fluid velocity.
In water treatment practice, orthokinetic flocculation is the most predominant mechanism in the aggregation of drag coefficient (Ives 1978b). Therefore the pow~r is given by: where Cd • Drag coefficient The expression has the f~llowing form (Parker et al. 1972): parameters that need to be optimized are: (1) Coagulant dosage; (2) coagulation pH; (3) mixing intensity; and (4) mixing time. 27 The basic laboratory tool used for determining the above parameters in water treatment operations is the jar test. A stirrer with flat paddles has traditionally been used as the impeller in jar testing. There has been a great deal of work to improve the jar test procedure by investigator's such as Camp (1955Camp ( , 1968), Harris et al. With the use of the jar test procedure, there are many techniques that have been used as performance indicators.
The use of particle counting in full scale treatment plants has been adopted as a supplement to the turbidimeter. This is because there are no specific criteria being set on the number of particles in the finished water as well as the 30 nonexistence of a universal relationship between particle count and turbidity. Therefore, particle counting is used by researchers as a tool in understanding the actual changes that take place during the different treatment processes.

Granular-Media Filtration
Filtration is the second most important water treatment, the first being disinfection.  Transport by inertial impingement is the predominant removal mechanism in fibrous filters used to clean air, but this mechanism is not of great significance in rapid sand filtration owing to the high viscosity of water. Inertial impingement is the result of the attachment of particles that have sufficient momentum to move downward, deviating from the original path around the sand grain, resulting in impingement on the sand grains.
hence Transport by the interception and chance mechanisms have been suggested to occur when the separating distance between the moving particle and the sand grain is less than or equal to half the diameter of the particle. In water filtration, most of the suspended particles have a higher density than water and because the fluid velocity near the media grains approaches zero, therefore transport by sedimentation would be expected to occur. 35 The transport of particles to the grain media can be expressed in term of the single collector efficiency of a media grain which is defined as the ratio of the rate at which suspended particles collide with a media grain to the rate at which they flow towards it. Considering only the transport by diffusion, interception and sedimentation, the total single collector efficiency,llT, of a media grain is written as (Yao et al. 1971 andO'Melia 1985): In terms of the sand size, particle removal and head loss are inversely proportional to the diameter of the media. The media size determines the degree of suspended The effect of media shape on the process of direct filtration has not been . investigated before. However, it has been shown that for the same size of coal and sand, the former has a higher storage capacity due to the greater bed porosity. The porosity of a clean filter bed is normally 42-43 per cent for rounded sand and about 52-55 per cent for anthracite (Hudson 1969). The media shape is expected to influence the pore space and surface area of the media that is available for suspended particle attachment. In section 2.6, the influence of influent suspension particle size distribution on the performance of granular media filtration will be presented.
The pH of the influent suspension influences both the effluent quality and head loss . Hutchison (1976) and Hudson et al. (1967) showed that the filter effluent turbidity rose as the pH of the flocculated water was increased above 7.00. Tbe run time was also increased as the pH increased above the value of 7.00.

Particle Size and Distribution
Particle counting is a useful measurement technique permitting the quantification of suspended matter as well as providing a better understanding of the changes in particle size distribution in the water treatment processes.
Basically, there are three methods that are commonly used in particle counting. These are: (1) Manual via microscopic examination; (2) electrical resistance; and Particle counting by light blockage is the most commonly used technique in the water treatment applications.
The technique is based on the principle of light blockage.
Particles in fluid suspensions flow through a channel past a window of known area which has a collimated light beam that shines through the fluid at right angles to the direction of flow. When a particle passes across the beam, it partially blocks the light falling on a photodiode. The pulse generated is proportional to the projected area of the particle, and particle size is specified in terms of equivalent spherical diameter. The most commonly used instrument is the HIAC Model PC-320 . It is easy to operate and capable of providing reproducible results compared to other instruments. The resulting particle counts and sizes are comparable to those achieved by electronic particle counting and sizing (Treweek et al. 1977).
Natural waters contain many particles of various sizes.
The distribution of these particles can be described mathematically. This particle size distribution may be the most important physical characteristics of the system. It has been shown that the particulate size frequency distribution in natural waters follows closely the power-law Values of for natural waters ranged from 1.80 to 4.50 as was reported by Kavanaugh et al. (1980). The exponent ~ will be referred to as the distribution slope from now on.
Particle size distribution can be presented in three forms depending on the method used to describe particle size. Particle volume, surface area, or diameter may be used, resulting in the following particle size distribution . The influence of mixing intensity and time on the particle size distribution has not received great attention.
The influence of these operational parameters on the The results were shown in plots of the number of particles 49 against the particle size for raw water, coagulated-flocculated water, and filtered water. There was no attempt to use the distribution slope, ~ as a performance indicator. The purpose of the study was to investigate the optimum mixing time prior to direct filtration using particle size analysis. By studying the changes in the particle size distribution resulting from the changes in the mixing intensity and time, it could be easier to understand the influence of these parameters on the filtratidn process in direct filtration.
The particle size distribution in the filter effluent of full scale or even pilot scale treatment plants has never been investigated in full detail as a function of the influent particle size distribution. Traditionally, the effluent quality might have been reported in terms of the total particle count with reference to the size range of measurements. Some treatments plants Monscvitz et al. 1983) use 20 particles/ml as the maximum in the effluent of the filter, others use a particles/ml (Tate et al.

1977)
and some use so particles/ml (Tate et al. 1978 This section deals with the influence of the influent suspension particle size distribution on the performance of granular media filtration, specifically the effluent quality and particle size distribution, and head loss development.
since no previous comprehensive work was preformed in this area, then the discussion will be based on the available related literature as well as the author's own hypothesis.
The discussion will be focused mainly on the influence -of the pretreatment coagulant dosage, mixing intensity and time on the performance of granular media filtration.
There are many adapted to measure particles and the empirical parameters that have been the interaction between the suspended filter media. This is because the variables affecting particle transport and attachment are numerous and solid clarification is difficult. The first of these empirical parameters is the filter coefficient developed by Iwasaki in 1937. This coefficient is nothing more than a proportionality constant relating the removal of solids to the filter depth. Other parameters that have been used in the past are the bulking factor (which is related to the porosity of the deposited solids), the Filterability Number developed by Ives (1978a) and the specific deposit.
The Filterability Number is a dimensionless parameter that relates the filterability of a suspension to filter material taking into account clarification, clogging and flow rate. 51 The specific deposit is the volume of retained particles in a unit of filtering material.
As was mentioned before, the particle removal is inversely proportional to the diameter of the media, and regardless of the media size, there are always particles in the effluent which pass through or are dislodged from the filter. With fine media, these particles are in the small size range. Therefore, it would be expected that the particle size distribution will be sensitive to a slight change in the number of particles at any size within that range, hence the distribution slope, ~, will be more scattered as would be the case when coarser media is used.
In cases where the filter run is controlled by particle breakthrough, the distribution slope, ~, is expected to

Influence of Coagulant Dosage
As the coagulant dosage is increased, more particles are destablized and agglomorated into larger particles.
Since there will be an increase in the number of larger sized particles, the distribution slope, will be shifted to a lower value. Assuming no restabilization of the particles will occur, at least at the concentrations used in this study, then the effluent quality is expected to improve as the coagulant dosage is increased. However, due to the rapid filling of the pores within the bed resulting from the increase in the influent solids, the total volume of water produced will decrease dramatically and the effluent quality 53 will also deteriorate. Therefore, the use of high coagulant dosage would result in low effluent quality as well as less volume production as compared to other dosages.
The deterioration of the effluent quality could be the result of the increase in the flow velocity within the bed as the pores fill since the flow rate is kept constant during the run. Therefore, some of the materials that have been previously retained by the media will appear in the effluent as the flow rate is increased. The effluent distribution slope,~ , is expected to follow the same trend as that of the influent. As the coagulant dose increases, the slope of the distribution function will decrease for both the influent and the effluent. However, when the coagulant dose is increased to a level that causes a rapid clogging of the bed then the particles that are released from the bed will cause the effluent distribution slope to deviate from this pattern.
The direction of the deviation is not known. Perhaps the magnitude of the distribution slope will give an idea of the type of particles that have the tendency to be detached from the sand grains.
An increase in the coagulant increase in the solids produced Therefore, the rate of head loss increase in the chemical dosage.
dosage results in an from a given raw water.
will increase with the The relationship between the chemical dosage and the filter run length has been shown to be of the exponential type (Shea et al. 1971, Hutchison et al. 1974, Hutchison 1976and Foley 1980. In terms of the distribution slope, ~ , the higher the coagulant dosage the larger the floes, hence the lower the ~ value. As the size of the floes increases, the chance of non-floe particle penetration is reduced and suspended solids penetration of the filter is minimized. This effect is more dominant with SS the fine to medium sand.

Influence of Mixing Intensity
The mixing intensity has a great influence on the particle size distribution of raw waters. High mixing intensities produce floes that are small in size resulting in high distribution slope values. When the · mixing intensity is low, the distribution of the flocculated particles is slanted towards the larger sizes and hence less ' filter penetration is expected.
Most of the floes will deposit on the surface and they will aid in the removal of the smaller particles. However, it is expected that most of the effluent particles will be in the small size range since deep penetration is not expected at low mixing intensities.
The effluent distribution slope, ~ , is expected to be inversely proportional to the mixing intensity. At a low mixing intensity the remaining small particles will go through the bed and show up in the effluent resulting in a high value for the effluent distribution slope. When the mixing intensity is high, small particles will be produced in the flocculator and more particles with a wide range of sizes will · show up in the effluent resulting in a low effluent distribution slope. It should be noted that the media size plays an important role in the effluent distribution slope and the above hypothesis is a generalized one.
Perhaps there is an optimum mixing intensity for different media sizes that will give the best particle 56 removal.
The It is expected that the high mixing times produce fractions of the actual particles whereas high mixing intensity produces whole small particles. Therefore, the degree of penetration and arrangement of the particles within the bed are quite different and the effect on the effluent quality and head loss also will be different. 57 The effluent quality in terms of the total particle count should improve with an increase in the mixing time beyond the optimum value. This is because smaller particles penetrate deeper in the sand bed and the removal of subsequent particles would be improved due to particle-particle removal mechanism. The optimum mixing time causes large floes to be deposited on top of the sand bed and hence smaller particles will pass and show up in the effluent. This phenomenon should be more apparent using fine medJ.a as compared to coarser media. Also because of the rapid clogging of the sand bed, it is possible that some of the attached particles will be detached by the high flow through the pores causing an increase in the total particle count in the effluent water. Bed clogging is considered either surface or pores of the media depending on the media size.  greater than or equal to the lower limit of that interval.
The instrument was calibrated at the factory by using monosized spheres with known sizes.

Particles
The particles selected for use in this study were a colloidal clay product obtained from Pennsylvania Glass Sand Corp., Pittsburgh, PA.
The particle characterization of this material is shown in           14.
The tubes were connected below each column to 1/4 inch flexible tubing that were connected to the piezometer board.
Each tubing was connected to glass tubings that extended the whole length of the piezometer board. The piezometer board The sand media used in the three columns had the following•sieve analysis: Filter No.

Sand preparation
The sand was received in 100 lb sacks that contained a mixture of different grain sizes. Therefore, it was necessary to sieve the sand to obtain the desired grain sizes. This was done by hand sieving 100 grams at a time using two sieves. One to pass all the grains and the other to retain the size needed. The sand then was washed with ' tap water to get rid of all the dust and foreign objects, then was allowed to dry in the oven at 70°F. The dry sand then was placed in the appropriate column so that the depth of media was about 5 inches above the required depth of 17 inches. The sand bed was then backwashed for 1 hour.
Backwashing resulted in the expansion of the sand bed so that fine and less dense materials would rest on top of the heavier media. The sand bed was allowed to settle very slowly in the column, after which the top 5 inches were scraped off and discarded. The 5 inches of sand that were scraped off comprised the fine media that did not pass through the bottom sieve during sieving, as well as the media that was not sand but had a lower density with the same size as the sand. 78

Experimental procedure
The experimental work was divided into three phases.  Three problems were encountered at the start of the experimental investigation.
The first was the dilution water that was needed to dilute the samples for particle analysis.
The deionized water available in the laboratory had particle counts that were too high for the purpose of dilution (about 20 particles/ml). Methods to reduce this particle count were not successful. One method involved filtering the water continously through a 0.45 micrometer membrane filter in a closed system using a peristaltic pump.
The particle count increased due to the abrasion of the pump against the plastic tube. The problem of high particle counts was finally solved by allowing the deionized water to continuously drip for a few hours then collecting the volume needed in a container that was washed with acid. This method resulted in a particle count of less than 4 particles/ml. 81 The second problem was the increase in the number of particle counts after the pH adjustment.
adjusted by using calcium hydroxide (lime).

The pH was
It happened that the lime had particles that were causing an increase in the particle count. Switching to sodium hydroxide solved the problem. Particle counts did not increase after the addition of sodium hydroxide. The three values were then totaled and were averaged for plotting purposes.
The total number of particles was plotted against the dilution ratio on a log-log paper as shown in Figure 3.9.
The minimum dilution was that value corresponded to the intersection of the tangent line to the negatively sloped 83 00 ""   portion o~ the curve.
In Figure 3.9 the minimum dilution ratio was about 16.0 (equivalent to 6.25 mg/L of mass concentration). Any dilution ratio equal to or above that value would result in an acceptable particle count. In all the experiments the dilution ratio was set at 40 (equivalent to 2.5 mg/L of mass concentration). This ratio was selected in order to minimize the withdrawal of large samples during the experimentation and also to avoid a particle count that is near the deionized water particle count. Also with tnis dilution, ratio it was possible to collect many samples without affecting the total volume within the mixing unit.
In all of the six experiments, the temperature (12°C) as well as the raw water particle concentrations were kept     93 In all of the 24 runs, the solution pH, temperature and raw water particle count were kept constant within a narrow range.
The pH was kept between 6.90 and 7.15, the temperature between 12 and 14° C and the total particle count had an average of 16,700 particles/ml. The university tap water was analyzed daily for ,particle counts and the results are shown in Table 3 Figure 3.3).
The chemical requirements were set according to the total flow rate, and the water temperature was set by adjusting the cold and hot water valves.
Since the suspension was rapidly mixed, it was necessary to investigate the influence of rapid mixing on the particle size distribution of the coagulated water since with different detention times, the flow rate was different.
This resulted in different detention times in the rapid mixing unit which could influence the rate of initial flocculation.
To accomplish this, two flow rates were selected that would give different detention times. These Prior to the start of filtration, the system was allowed to stabilize for at least 3 hours. This was to ensure a constant pH, temperature, water particle count and desired mixing intensity.
97 Table 3 Figure 3.12 Velocity gradient versus rpm for the rapid mixing unit. 99 The filter run termination was based on two performance parameters, the effluent quality and the head loss. When the head loss across the entire sand bed reached the available head ( 7 • 5 ft), the filter run was terminated.
This was the case with the fine and medium sand. With coarse media the effluent quality was used as the controlling parameter for the termination, because with such a large media size, the head loss would never reach the available head before particle breakthrough. When the effluent • reached 10% of the raw water total particle count, the filter run was then terminated. The choice of the 10% limit was based on two factors. The first was the fact that this particle count was so low that dilution was not higher than what is used in practice. The reason for higher backwashing rates was to return the filter bed to its original state so comparisons can be made. Before the termination of backwashing, samples were taking for particle analysis to make sure that the sand bed was free from previously deposited particles. Closure of the backwash valve was done in intervals so that the sand bed would return to its original configuration. 101

RESULTS AND DISCUSSION
The results and discussion are divided into two main sections.
In the first section, the results of the pretreatment (coagulation/flocculation) experiments will be discussed.
In the second section, the results of the pilot plant experiments will be presented and discussed.

Pretreatment
The results to be presented in this section are those that were obtained from the six jar test experiments as well as those obtained from the four pilot plant mixing tank experiments. The discussion will be focused on the change of the total number of particles as well as the change in the distribution slope,~ , with different operational parameters.

Jar Test Results
The purpose of the jar test experiments was to study the effect of mixing intensity, coagulant dosage, pH, and mixing time on both the number of particles remaining in solution as well as the particle size distribution. 102 The first experiment was used to define and quantify the standard conditions. The data that were collected from all the six jar test experiments are listed in Appendix B.
The raw data were fitted to first and second order polynomial functions. The best fit was found to be a second order polynomial function.
The rationale for this kind of fitting as well as a discussion of the ~tatistics that compare the two fittings will be presented in the following   However, the rate of flocculation was not high enough to reach a minimum particle count at a short mixing time. This is attributed to the fact that the ratio of the coagulant 104 " I   As can be seen from Table 4.1, the second order fitting was much closer than the first order.
However, looking at the relationship between the total number of particles with respect to mixing time, it seems that the number of particles starts to increase after a certain mixing time. This is shown in             In practice, the optimum pH range for the removal of turbidity by alum flocculation is generally 6.00-7.80 (Kawamura 1976, Edwards et al. 1985. In Figure 4.10, the rate of flocculation increases sharply with an increase in the pH. For pH 7.00, and 8.00 the rate of flocculation is almost identical. At pH 9.00 the rate of flocculation is similar to pH 7.00 but the total rema~ning particles are lower which indicates that flocculation is continuing at this high pH. This is because different clay suspensions exhibit ~uite different pH optima and probably the clay that has been used in this study has a wide optimum pH range. If this is the case then the adjustment of pH above a certain value is not an important factor for removal of these type of particles. In contrast, kaolinite suspensions usually have a very sharp optimal pH point for turbidity removal by alum flocculation. Optimum pH is considered that value which would result in the minimum remaining particle count and beyond that value the number stays the same or very close for the same coagulant dosage. In In Table B         The time required to cause any change in the number of particles of any certain size is much longer at pH 6.0 than at 7.0. This is shown by the small change in the slope of the distribution function (Figure 4.14). At pH values below the optimum, the time required to notice any increase in the large size particles or a decrease in the smaller size is usually three times that when flocculating at the optimum pH values (Hannah et al. 1967).   .,  terms of the remaining total number of particles. This is shown in Figure 4.18 which was plotted the same way as                     (Table 4.7). Therefore up to the optimum mixing time, the data fit a first order whereas beyond that time the best fit is a second order polynomial function.
The following points summarize the above discussion:  The three operational parameters that were used to change the particle size distribution of the raw water were the coagulant dosage, mixing intensity and time. The influence of the particle size distribution produced by each of these three parameters on the filtered water quality and PSD will be discussed. The data generated from all of the 24 runs are listed in Appendix c.
The data that were obtained from the pilot plant were used to calculate the filter coefficient, specific deposit, and the Filterability Number. The parameters are required in order to be able to compare the effect of the different influent suspensions on the effluent quality. However, due to the fact that the ratio of the influent to the effluent particle count is very high, those parameters did not prove 154 to be useful tools for comparison purposes. As an example, the influent particle count was about 16,000 particles/ml and with fine media (0.35 mm), the effluent particle count was less than 10 particles/ml in most of the runs independent of influent conditions. Therefore, no obvious trend in the data were observed with respect to the above three parameters.
Due to these problems, qualitative comparisons and observations will be presented.    As the sand grain diameter increases, the ~ffluent quality deteriorates. This is shown in Figure 4.25 for filter number 2 (0.50 mm). Only when no alum was added did the effluent reach 10% of the influent total particle count .

I.nfluence of Coagulant Dosage
The other alum dosages caused the filter to be clogged before 10% was reached. Therefore, the run termination was  For filter number 2, only 8.0 mg/L of alum gave very good effluent quality for more than 20 hours. Below the 8.0 mg/L dose, the effluent quality was poor and above that dose the rate of deterioration was very high. Therefore, only the optimum dosage seem to give the best results with medium size media.
Using coarse media (filter number 3, sand size 1.00 mm) the effluent quality was not as good as that produced by fine and medium sand. The minimum particle count was about 162 °' w         This indicates that at a certain coagulant dose and beyond, the distribution of particles in the effluent is the same.
So overdosing will not only result in a short filter run but also will give the same PSD as when the optimum alum dose is used.

Influence of Mixing Intensity
The mixing intensity influences the particle size distribution in that it determines the largest floe size that can be formed.
As the mixing intensity increases, smaller particles are formed giving a higher distribution slope value. Higher mixing intensities produce smaller floes that tend to penetrate deeper in the bed and cause some particles to show up in the effluent. Because of particle penetration, particle to particle removal mechanism is reduced especially at the surface of the media.     indicates that mixing intensity has less effect on effluent quality than the chemical dosage, at least with fine to medium sized sand media. However, the filter run length seems to be directly proportional to the mixing intensity.
The runs were terminated because the head loss reached the available head in both filter number l and 2 and not because of the effluent quality.
When the filter run length is controlled by the effluent particle count as was the case for filter number 3, the filter run length was inversely proportional to the mixing intensity. This shows the relationship between the media size and the particle size in term of the removal mechanisms. It indicates that straining is an important 178 removal mechanism.
The influent distribution slope, ~ is expected to have a direct relationship with the mixing intensity. ....     possibility is there could be a relationship between the particle size distribution, which is a function of the mixing intensity, and the media size. Therefore, an optimum particle size distribution could exists for any specific media size.
The optimum effluent particle size distribution is the one that contains the least number and smallest size of particles. This is because large particles (greater than 15 microns in size) have been suggested as the major cause of interference with the disinfection process (Beard et al. 1977).
In terms of the number of particles in the effluent, both mixing intensities (25 and 100 s-1) gave close total particle counts for fine sand media. As the media size increases, the influence of the mixing intensity on the effluent particle size distribution does not show any significant difference between the high and low mixing intensities that were used in the study. In with coarse media, the effluent particle size distribution did not show great dependence on the influent particle size distribution except at the early filtration times. was not as rapid.    Table   4 • 9) • To avoid effluent rapid deterioration and to provide the best maximum volume output, medium mixing times are desirable with medium sand size filters.
Coarse media (Figure 4.40) gave a lower effluent total particle count for the 15 minutes than the 30 minutes mixing time.
The region of constant effluent particle count increased but the deterioration was as rapid as before.
With a low mixing time, the floes that have been formed have  Compared to the average total particle count achieved in treatment plants, the fine sand media yielded a better effluent particle count with all mixing times for almost the whole length of the run. The medium sand also gave an effluent total particle count below the average with all three mixing times but not during the whole run. The coarse media was above the average (10 particles/ml) for all mixing times.
The distribution slope increases with the increase in the mixing time.

Influence of Influent Particle Size Distribution on
Head Loss The influence of the influent particle size distribution produced by the three operational parameters on the head loss at different depths will be presented and discussed in this section.

4.2.2.l Influence of Coagulant Dosage
The tbtal head loss across the bed is plotted against run time for four different alum dosages. This is shown in  ....     .....     As the media size increases, the influence of high mixing time becomes less significant in terms of the head loss. This can be seen by comparing Figures D-7 and D-9.
In Figure D    ......   5. The effluent quality and particle size distribution 219 from the filters packed with smaller diameter filter media proved to be more sensitive to changes in the influent particle size distribution than did filters with coarser media.
6. The relationship between the rate of head loss development and the influent particle size distribution was found to be an exponential function for both fine and medium filter media and linear for coarse media. 220

RECOMMENDATIONS
Further research is needed in the area of particle size distribution as related to water treatment practices. More specific recommendations for future research in this area includes the following: