Plate 1: Washington Aqueduct, D. C. Filtration Plant Injector Sand-Handling System

AMERICAN SOCIETY OF CIVIL ENGINEERS

INSTITUTED 1852

TRANSACTIONS

Paper No. 1191

WATER PURIFICATION PLANT, WASHINGTON, D. C.
RESULTS OF OPERATION.¹

By E. D. Hardy, M. Am. Soc. C. E.

With Discussion by Messrs. Allen Hazen, George A. Johnson,
Morris Knowles, George C. Whipple, F. F. Longley, and E. D. Hardy.

The Washington filtration plant has already been fully described.² At the time that paper was written (November, 1906), the filtration plant had been in operation for only about 1 year. It has now been in continuous operation for 5 years, and many data on the cost, efficiency, and methods of operation, have accumulated in the various records and books which have been kept. It is thought that a brief review of the results, and a summary of the records in tabular form, will be of interest to the members of the Society, and it is also hoped that the discussion of this paper will bring out the comparative results of operation of other filter plants. As a matter of convenience, the following general description of the plant is given.

Description of the Filtration Plant.—The Washington filtration plant was completed and put in operation in October, 1905. It consists of a pumping station for raising the water from the McMillan Park Reservoir to the filter beds; 29 filters of the slow sand type, having an effective area of 1 acre each; the filtered-water reservoir, having a capacity of about 15,000,000 gal.; and the necessary piping and valves for carrying water, controlling rates of filtration, etc.

In the pumping station, there are three centrifugal pumps, which are directly connected to tandem compound engines; two sand‑washer pumps; three small electric generating sets for furnishing electric light; and four 200‑h.p., water-tube boilers.

Each of the centrifugal pumps has a nominal capacity of 40,000,000 gal. per day when pumping against a head of 21 ft., and each sand‑washer pump has a capacity of 2,500,000 gal. when pumping against a head of 250 ft. The electric light engines and generators supply the current for lighting the pumping station, the office and laboratory and other buildings, and also the courts and interior of the filter beds, and for operating a machine‑shop.

The filters and filtered‑water reservoir are built entirely of concrete masonry. The floors are of inverted groined arches on which rest the piers for supporting the groined arch vaulting. All this concrete work is similar to that in the Albany, Philadelphia, and Pittsburg filters.

The filters contain, on an average, 40 in. of filter sand and 12 in. of filter gravel. The gravel is graded from coarse to fine; the lower and coarser part acts as part of the under‑drain system, and the upper and finest layer supports the filter sand. The raw water from the pumps is carried to the filters through riveted steel rising mains which have 20‑in. cast‑iron branches for supplying the individual filters. The filtered water is collected in the under‑drainage system of the several filter beds, and is carried through 20‑in., cast‑iron pipes to the regulator‑houses. These regulator‑houses contain the necessary valves, registering apparatus, etc., for regulating the rate of filtration, showing the loss of head, shutting down a filter, filling a filter with filtered water from the under‑drains, and for turning the water back into the raw‑water reservoir, or wasting it into the sewer. From the regulator‑houses, the filtered water flows directly to thefiltered‑water reservoir. Generally, five filters are controlled from one house, but there are two cases where the regulator‑houses are smaller, and only two filters are controlled from each.

The dirty sand removed from the filters is carried by a portable ejector through one or more lengths of 3‑in. hose and a fixed line of 4‑in. pipe, to the sand washers. From the sand washers, the washed sand is carried to the reinforced concrete storage bins, each of which has a capacity of 250 cu. yd., and is at such an elevation that carts may be driven under it and loaded through a gate.

Until April, 1909, the sand was replaced in the filters by carts which were filled through the gates in the sand bins. It was then hauled to the top of the filter beds and dumped through the manholes on the chutes, which could be revolved in any direction. These chutes were used to prevent the sand from being unduly compacted in the vicinity of the manholes, and to facilitate spreading it in the filters. Since April, 1909, all the sand has been replaced by the hydraulic method. An ejector is placed under the gate in the sand bin, and the sand is carried in a reverse direction from the bin through the 4‑in. piping and one or more lengths of hose to the filter bed. This process has lowered the cost of re‑sanding considerably, and present indications are that it will prove entirely satisfactory in every way.

The average effective size and uniformity coefficient of the filters are shown in Table 1.

Description of Washington Aqueduct.—The water supply of Washington is taken from the Potomac River, at Great Falls, about 16 miles above the city. At that place, a dam has been built across the river, which holds the water at an elevation of 150.5 ft. above mean tide at Washington. From Great Falls the water flows by gravity for a distance of 16 miles through a 9‑ft. conduit, three reservoirs, and a tunnel. From McMillan Park Reservoir, the last of the three, the water is lifted by centrifugal pumps about 21 ft. to the filters. After passing through the filters, it flows to the filtered‑water reservoir, and later to the city mains. In its passage from Great Falls to the filters, the water flows through three settling reservoirs, which have already been referred to. These reservoirs are known as the Dalecarlia, the Georgetown, and the McMillan Park Reservoirs, and have available capacities of 141,000,000, 140,000,000, and 180,000,000 gal., respectively.

Turbidity.—The Potomac River water is rather turbid, the turbidity being caused by very fine particles of clay. The river is subject to sudden fluctuations, it being no uncommon thing to have a turbidity of 100 one day, and 1,000 the next. The high turbidity usually disappears about as rapidly as it comes, and is seldom higher than 500 for more than 5 days at a time. It is frequently the case, however, that a succession of waves of high turbidity will appear so close together that the effect of one has not disappeared before that of another is felt.

The clarification of the water supply begins at the dam at Great Falls. Here it is a clarification by exclusion, for when an excessive quantity of mud appears in the river water, the gates are closed, and the muddy water is allowed to flow over the dam and form mud‑bars in the Lower Potomac, while the city is supplied from the water stored in the three settling reservoirs. Until a comparatively recent date, the excessively muddy water was never excluded, having been taken, like other decrees of Providence, as it came.

During the summer of 1907, the practice of shutting out water with a turbidity of 500 or more was established for the warm months. This practice was discontinued during the cold months, as it was feared that a very high consumption of water might occur at the time of low water in the reservoirs, and so cause a partial famine. During the winter of 1909‑10, however, the gates were closed, as was the practice throughout the summer months.

When the reservoirs are well filled, and the consumption of water is less than 70,000,000 gal. per day, it is safe to close the gates at Great Falls for a period of about 4 days.

Figure 1—Plan and Profile of Washington Aqueduct.

While a considerable reduction in turbidity is effected in each of the reservoirs, the bulk of the mud is deposited at the upper end of Dalecarlia Reservoir. This reservoir had become so completely filled, that, in 1905, it was necessary to dredge a channel through the deposit, in order to allow the water to pass it. During the summers of 1907 and 1908, a 10‑in. hydraulic dredge removed more than 100,000 cu. yd. of mud which had been deposited in this reservoir. The mud deposited in Georgetown and McMillan Park Reservoirs is so fine that the accumulation of many years is not very noticeable in its effect on the depth of water.

The particles of clay which remain in the water after its passage through the three reservoirs, are so exceedingly small that they do not settle out in any reasonable length of time. Even the filtration of the water through one or more slow sand filters occasionally fails to remove the last trace of turbidity. This is especially true in the colder months, and not a winter has passed when the water supply has not been noticeably turbid at some time.

A general idea of the quantity of mud contained in the river water, the quantity excluded by closing the gates at Great Falls, and that removed by sedimentation and filtration, may be gained from Table 2, which is, of course, only a rough approximation.

Table 2 also shows that the gates were closed 10.50% of the time, thereby excluding 40.06% of the total suspended matter which otherwise would have entered the system.

The turbidities, bacterial counts, and chemical analyses of numerous samples of water are shown in Tables 3, 4, 5, and 6. The amount of work done in the pumping station, average consumption of water, death rate from typhoid fever, and filter runs are shown in Tables 7, 8, 9, and 10.

Raking.—At the time the filters were first put in service, the sand bins had not been completed, and, consequently, the work of cleaning the filters was carried on in the old‑fashioned way of scraping by hand and wheeling out the sand in barrows. This method of cleaning was used from October, 1905, to April, 1906; then the regular sand‑handling system was commenced.

At times, during the first two summers the filters were in operation, considerable difficulty was experienced in keeping them cleaned as fast as was necessary to provide an ample supply of filtered water. For a short period in each summer it was found necessary to organize night shifts, and keep the work of cleaning in progress for from 16 to 24 hours per day.

Figure 2—General Plan of Washington Filtration Plant Showing Finished Surfaces.

In order to relieve the situation at such times, the expedient of raking was tried. This was first attempted with the filters filled with water; the effluent was first shut off in order to prevent a downward flow of water, and the filter was then raked or harrowed from boats. This method was not satisfactory, however, as the work was neither as uniform nor as thorough as necessary. Later, the filters were drained to the necessary depth, and the surface of the sand was thoroughly stirred with iron garden rakes. The filters were then filled with filtered water through the under‑drains and put in service.

This latter method proved so satisfactory that it has been resorted to at all times when the work was at all pressing. When the runs were of short duration, and the depth to which the mud had penetrated the filter sand was slight, a raking seemed to be nearly as effective in restoring the filter capacity as a scraping; it could be done in 8 hours by 3 laborers, and there seemed to be no ill effects from lowered efficiency.

The length of runs, depth of scraping, etc., after the scraping or raking, are shown in Tables 10 and 11.

Sand Handling.—For the first three years of operation, the sand was carried from the sand bins in carts and dumped through the numerous manholes of the filters on chutes which could be revolved in various directions, in order to facilitate the spreading of the sand evenly over the surface of the filter.

About a year ago, however, this method was changed, by substituting sand ejectors for the carts. By this method, an ejector is either attached to, or placed directly under, the outlet gate of the sand bin, the gate is opened, and the ejector is started. From this ejector, the sand is carried back through the line of 4‑in. fixed pipe, and one or more lengths of 3‑in. hose, to the point of discharge in the filter bed which is being re‑sanded.

Figure 3—Washington Aqueduct, D. C., Filtration Plant. Sand Handling, System.

Figure 4—Washington Aqueduct, D. C., Filtration Plant. Washer Sand‑Handling, System.

Figure 5—Washington Aqueduct, D. C., Filtration Plan. Ejector Sand‑Handling, System.

In re‑sanding a filter, it is first filled with water to the proposed depth of the sand layer. The outlet end of the hose is connected to a 3‑in. pipe which is supported on a boat, and the sand is discharged through this pipe at the point required. Work is first begun at the far end of the filter, and it is gradually filled by swinging the boat from side to side and backing it by degrees to the front end.

At first it was feared that a small quantity of mud would be deposited on the surface of the old sand, and that this mud would ultimately cause subsurface clogging. For this reason, when this method was first adopted, a man was required to rake the sand very thoroughly in front of the discharge. Later, it was found that by giving the end of the discharge pipe a slope of about 45° downward from the horizontal, the force of the current of sand and water could be depended on to cut the old surface of sand to any required depth, and move it ahead together with the new sand, thus completely breaking up the possible mud layer between the old and new sand layers. After having used this method almost exclusively for 15 months, in which time eleven filters have been re‑sanded, and 24,531 cu. yd. of sand have been replaced, there seems to be no indication of an increased initial loss of head. The sand is very compact, and has no apparent tendency to separate into different sizes. The general appearance is similar to that of very fine sand on the seashore. The filters re‑sanded in this way have been considerably more efficient than those in which the sand was replaced with carts, and as yet, no harmful results have been noted. The rate at which the sand is replaced is shown in Table  12, and the cost of labor for sand handling is given in detail in Table 14, which shows that quite a perceptible saving has been effected by the hydraulic method.

The figures showing the cost for sand handling do not include any charge for the quantity of water used, that item having been carried on the pumping‑station account.

The cost for pumping water for sand handling, including all labor, materials, and repairs, amounts to $0.06 per cu. yd. of sand ejected and washed, and $0.03 per cu. yd. for replacing.

In addition to the water used for carrying the sand which is being replaced, it is customary to keep a slight upward flow in the filter, thus using about 500,000 gal. of filtered water per day for this purpose. Assuming the value of this water to be the total cost for pumping, filtering, etc., or $3.80 per 1,000,000 gal., the cost per cubic yard of sand replaced would be about $0.02 when one ejector is used, and $0.01 when two are in operation.

It is not considered absolutely necessary to have an upward flow of water in the filter which is being re‑sanded, and it is not always done. It was used, however, as an additional safeguard against the formation of a stratum of mud between the old and new layers of sand while the hydraulic method was in an experimental stage.

The quantities of sand removed from the filters per scraping and the rates of sand handling are shown in Tables 11 and 12.

Cost of Operation.—It is frequently difficult to compare the relative cost of corresponding items for different plants, because of the different methods of dividing the cost and the varying opinions of the officials as to what should properly be charged to each item.

In order that the data may be in sufficient detail to permit it to be rearranged to compare with other plants, a list of employees and charges for supplies is given in Table 13. This list accounts for the entire appropriation for the care and maintenance of the filtration plant, including pumping the water to the filters, parking and caring for the grounds, buildings, roads, sidewalks, etc. The cost for the various items per million gallons pumped to the filters is shown in Table 14, and the cost per cubic yard of sand handled in Table 15.

Preliminary Treatment.—Before the present filtration plant was designed, Rudolph Hering, George W. Fuller, and Allen Hazen, Members, Am. Soc. C. E., made an investigation and report. This report was dated February 18th, 1901, and contained the following paragraph:

Notwithstanding this opinion, considerable prejudice existed among the citizens of Washington against the use of a coagulant, and, as finally passed, the bill providing for the construction of the filters did not include an appropriation for the coagulant.

The results obtained from operating the filters being such as to justify the conclusions in the report referred to, an experimental plant was constructed for the purpose of studying the efficiency of various methods of preliminary treatment of the water. This plant consisted of three cylindrical concrete filter tanks, each 10 ft. in diameter. These tanks were filled with the layers of gravel and sand necessary to make them represent as accurately as possible the large slow sand units of the main filtration plant. Means were also provided for giving a preliminary treatment to the water supplying each of these experimental slow sand filters. In two cases, the preliminary treatment was rapid filtration, while the third consisted of sedimentation and coagulation. The sedimentation tank was of sufficient size, when compared with the area of the experimental slow sand filter, to represent the Georgetown and McMillan Park Reservoirs when used in connection with the large filters. The first preliminary filter was very similar in construction and operation to a mechanical filter. The sand for this filter was taken from the main filters, and, consequently, was finer than is generally used in mechanical filters. The second preliminary filter was a Maignen scrubber. It consisted of a cylindrical concrete tank, 4 ft. in diameter and 8‑1/2 ft. deep, which contained 12 in. of cobble‑stones on the bottom, then, successively, 12 in. of egg‑size coke, 12 in. of stove‑size coke, 24 in. of nut‑size coke, and 24 in. of sponge clippings as the final or top layer.

The two preliminary filters were operated at a rate of about 50,000,000 gal. per acre per day, and the three slow sand filters at rates of from 3,000,000 to 4,000,000 gal. per day.

This plant was put in service during the early part of February, 1907, and was kept in practically continuous operation until the end of July, 1908.

Figure 6—Washington Aqueduct, D. C.
Experimental Filters Below Dalecarlia Reservoir Coagulating Basins and Apparatus.

For convenience in referring to the different systems, the combined rapid and slow sand filter will be designated as Filter Plant No. 1, the combined Maignen scrubber and slow sand filter as Filter Plant No. 2, and the combined coagulating basin and slow sand basin as Filter Plant No. 3.

The length of run of Filter Plant No. 1 was relatively long at first. The rapid rate of filtration, however, tended to carry the clay, which was suspended in the applied water, to a considerable depth in the filtering material, so that the runs gradually decreased in length until they were reduced to about three days. Unfortunately, it was necessary to use unfiltered water for washing, which, together with the great penetration from the applied water, finally made it necessary to remove all the filtering materials, and wash them.

Although this preliminary filter was operated at a high rate, its efficiency was quite satisfactory. In fact, at times when the applied water was comparatively good, very little work was left for the slow sand filter. At times of high turbidity, however, some of the exceedingly fine mud in the applied water passed through this filter, as well as the slow sand filter connected with it, and it proved to be absolutely impossible to produce a clear effluent at all times with this combination.

Filter Plant No. 2 proved more economical and convenient in operation, but somewhat less efficient than Filter Plant No. 1. Neither filter could be depended on to give a clear effluent when the applied water was turbid.

In the operation of Filter Plant No. 3, sulphate of alumina was used when the applied water contained too much turbidity to be treated satisfactorily by slow sand filters.

When the water was comparatively clear, either one of the three systems, or slow sand filtration alone, was entirely satisfactory. At times of high turbidity, however, Filter Plant No. 3 was the only one which could be depended on to produce a clear effluent.

A fair comparison between the results of the three systems when treating turbid water in January, 1908, is given in Table 16.

Table 16 shows very clearly that neither Filter Plant No. 1 nor No. 2 would prove at all satisfactory when treating turbid water, while No. 3 could be depended on under all conditions. The results of operation are shown in detail in Tables 17, 18, and 19. It will be noticed in Table 17, that on March 10th, 1908, Filter Plant No. 1 was put out of service and a Puech system of preliminary filters was substituted for it.

The Puech preliminary filters consisted of five units containing gravel of varying sizes through which the water was filtered successively before it was finally applied to the final slow sand filter. A general idea of this system may be obtained by referring to Figure 8.

It is unfortunate that this system was not in operation in January, 1908, when the water was cold and turbid. The results, however, indicate that it would be no more successful than either Filter Plant No. 1 or No. 2.

Experimental Rate Studies.—In September, 1908, an experimental plant consisting of six small filters was put in operation. The object of these experiments was to study the relative efficiencies and cost for the operation of slow sand filters when operated at different rates.

The units of the plant consisted of cylindrical galvanized‑iron tanks 4 ft. in diameter and 9 ft. high. The filter sand in these tanks was taken from the supply for the main filters. It was supported on gravel layers and supplied with under‑drains of suitable sizes for the proposed rate of flow in each case.

The units of the experimental plant were designated as Nos. 1, 2, 3, 4, 5, and 6, and it was the original intention to operate them at rates of 1,000,000, 3,000,000, 6,000,000, 10,000,000, 30,000,000, and 100,000,000 gal. per acre daily, respectively.

This schedule of rates was carried out in a general way with all the filters, with the exception of Nos. 5 and 6. For these, the rates were found to be higher than could be maintained for any great length of time, owing to the deeper penetration of the mud in the filter sand, which caused high initial losses of head, short runs, and deep scrapings. A rate of about 30,000,000 gal. was maintained in the case of Filter No. 5 from the time it was started on September 9th, 1908, until November 8th, 1909, when it was reduced to about 17,000,000 gal., which rate was maintained thereafter until the filter was shut down in February, 1910.

Figure 7—Plan of Filter‑House and Equipment.

In the case of Filter No. 6, it was found impossible to maintain a rate of 100,000,000 gal. for more than a very few days at a time. It was started at about this rate, however, at the beginning of each run, and kept as high as possible for the remainder of the time during the first seven runs. At the end of the seventh run, on October 17th, 1908, the filter was given a very deep scraping and re‑sanded.

Figure 8—Diagrammatic Sketch Showing Arrangements for Testing "Puech" System of Water Filtration at Washington D. C., U. S. A.

The layer of clean sand restored the original capacity, and the filter was operated as before, but with gradually decreasing rates until December, 1908, when the rate was reduced to about 40,000,000 gal. Even this lower rate was too high to be maintained without removing and replacing a large part of the sand. The rates, therefore, gradually decreased to about 23,000,000 gal. on March 13th, 1909, when the filter was again re‑sanded. After this re‑sanding the rate was reduced to about 20,000,000 gal., and the filter was operated at approximately that rate until it was again re‑sanded on November 13th, 1909, when the rate was again reduced to about 14,000,000 gal., which was maintained until the filter was put out of service on February 28th, 1910.

This experimental plant was in service from September, 1908, to the latter part of February, 1910, or for about 1‑1/2 years, and the leading results are summarized in Table 20.