Page 432 - Water and wastewater engineering

GRANULAR FILTRATION 11-5

In the mid-1980s, deep-bed, monomedia filters came into use. The filters are designed to
achieve higher loading rates while at the same time producing lower turbidity in the finished
2
3
water. They operate at loading rates up to 600 m /d · m (25 m/h).
When pretreatment of the water is by coagulant addition, flocculation, and clarification, the
filter is classified as conventional filtration. If pretreatment consists of coagulation and floccula-
tion but not clarification, the filtration process is called direct filtration. The process is called
in-line or contact filtration when coagulant addition but only incidental flocculation is used. In
some processes, coagulation is followed by two filtration steps: a roughing filter followed by
another filter.

11-3 FILTER MEDIA CHARACTERISTICS
Grain Size
The grain size, or perhaps more correctly the grain size distribution, affects both the hydraulic
performance of the filtration process and the efficiency of particle removal.
The size distribution, or variation, of a sample of granular material is determined by passing
the sample through a series of standard sieves (screens). One standard series is the U.S. Standard
Sieve Series. The U.S. Standard Sieve Series (Appendix B) is based on a sieve opening of 1 mm.
Sieves in the “fine series” stand successively in the ratio of (2) 0.25 to one another, the largest
opening in this series being 5.66 mm and the smallest 0.037 mm. All material that passes through
the smallest sieve opening in the series is caught in a pan that acts as the terminus of the series
(Fair and Geyer, 1954).
The grain size analysis begins by placing the sieve screens in ascending order with the
largest opening on top and the smallest opening on the bottom. A sand sample is placed on
the top sieve, and the stack is shaken for a prescribed amount of time. At the end of the shak-
ing period, the mass of material retained on each sieve is determined. The cumulative mass is
recorded and converted into percentages by mass equal to or less than the size of separation
of the overlying sieve. Then a cumulative frequency distribution is plotted. For many natural
granular materials, this curve approaches geometric normality. Logarithmic-probability paper,
therefore, results in an almost straight-line plot. This facilitates interpolation. The geometric
mean ( X g ) and geometric standard deviation ( S g ) are useful parameters of central tendency and
variation. Their magnitudes may be determined from the plot. The parameters most commonly
used, however, are the effective size, E, and the uniformity coefficient, U. The effective size is
the 10 percentile size, that is, the media grain diameter at which 10 percent of the media by
weight is smaller, d 10 . The uniformity coefficient is the ratio of the diameter of media at which
60 percent by weight is smaller to the 10 percentile sizes, d 60 / d 10 . Use of the 10 percentile was
suggested by Allen Hazen because he observed that resistance to the passage of water offered
by a bed of sand within which the grains are distributed homogeneously remains almost the
same, irrespective of size variation (up to a uniformity coefficient of about 5.0), provided that
the 10 percentile remains unchanged (Hazen, 1892). Use of the ratio of the 60 percentile to the
10 percentile as a measure of uniformity was suggested by Hazen because this ratio covered the
range in size of half the sand. * On the basis of logarithmic normality, the probability integral

*It would be logical to speak of this ratio as a coefficient of nonuniformity because the coefficient increases as the magnitude
of nonuniformity increases.