Performance of

Fine Sand Fluidized Bed Biological Filters

by

Dallas E. Weaver, Ph.D.

Scientific Hatcheries

8152 Evelyn Cr.

Huntington Beach, CA

92646

6/10/92

Abstract

Introduction

One major goal of aquacultural engineering research is a closed-cycle fish production system which can produce tonnage amounts of aquatic products without the use of large amounts of water or land resources. However, like all real-world engineering problems, this goal must be achieved at competitive costs.

Unlike the problems of waste removal associated with the production of terrestrial animals (poultry, cattle, pigs, etc.), the problem of separating the waste products from the growth environment of aquatic animals is non-trival. In other words, the aquacultural engineer must create a complete life support and waste treatment system if he wants to produce aquatic animals at high densities with limited water usage.

One of the critical unit operations in designing the life support system for aquatic animals is the soluble waste removal operation. A major component of the soluble waste produced by growing aquatic animals is ammonia. Hence, a unit operation which will remove, oxidize, or incorporate ammonia is required in all closed cycle aquaculture operations. The fine media fluidized biological bed filter can be viewed as primarily a soluble waste removal device. Operated aerobically, it will oxidize, with suitable bacterial species, most soluble organic compounds. Anaerobic operation will result in the reduction of nitrates and sulfates.

A medium to coarse media fluidized bed biological (FBB) filter has been used for ammonia removal at a steelhead facility (Owsley, 1989).

A fine media fluidized bed biological filter consists of a tank and a method for uniformly introducing the input water into the bottom of a tank which is partially filled with a fine media such as a very fine sand ( ie, 60 to 90 mesh silica). As the water flows up thru the tank, the sand is expanded to form a fluidized level. Mechanically, this fluidized sand behaves like a true high density fluid. It has no shear strength, and seeks it own level.

Looking closely at the individual sand grains, they appear to be in continual freefall through the water, hence the mass transfer capability between the liquid and the particle surface is excellent. With the very small size of the media, the surface area available for bacterial growth, per unit volume, is exceptionally large ( about 1 ha/m3 or 10 m2/liter). This large surface area per unit volume combined with high mass transfer rates for the liquid to the surface, create an excellent habitat for bacterial growth. These bacteria will then oxidize the waste products.

If the media is very uniform, the bed can become mixed in a time period much less that the reproductive time of the bacteria. The water passing thru the bed can be described as almost plug flow. This combination of mixed flow on the solid phase particles and their attached bacteria, with plug flow of the liquid phase can allow a biological filter to have a discharge concentration of a particular chemical below the minimum concentration required for bacterial survival as long as the input concentration is above the minimum (can be artificially increased). This unique capability of operating below the minimum feed concentration can be useful in certain aquaculture designs and situations where very high water quality is mandatory ( broodstock maturation and larval development).

With a fine media, assuming that the method of water injection is gentle enough, the bacteria are not sheered from the media by the flowing water. This lack of removal of the bacteria allows the filter to be operated with very long sludge retention times (SRT). It is the authors view that a long SRT system is more biologically stable.

A fine media fluidized bed filter, when operated with low loading rates per unit volume (less than about 0.5 kg of BOD -- including ammonia-- per cubic meter day), can be substrate (food) limited. Under these conditions, there is more surface area or bacterial habitat than there is food available for the bacteria. This mode of operation eliminates the microbiological competition for habitat, thereby allowing competition and selection of bacterial strains to be determined by the ability of the bacteria utilize the waste products at low concentrations. Under these low loading conditions, there is zero net production of biomass in the filter, no media loss, and no maintenance required.

Low loading rate operation allows the systems to handle pulse loading situations better than other types of biological filters. This can be visualized by observing that the system contains a large inventory of starving microorganisms. When presented with food, these starving organisms will be able to increase their consumption much more than well fed, fast growing organisms. Under these conditions, the system will self-select for organisms that will remove nutrients when available, thereby handling pulse loading situations.

When operating at higher volumetric loadings ( greater than about 1 kg of BOD per cubic meter day) there is a net growth of bacteria in the filter. This net growth appears as a sludge-sand particle fluidized layer on top of the sand media layer. As high loading rates continue, this sludge layer continues go grow at the expense of the sand layer and will overflow the filter unless removed or otherwise processed. If there are other suspended solids separation unit operations in the system, it is possible to introduce a sheering device in the sludge layer which will separate the biomass form the media, thereby allowing the media to return to the sand layer and the biomass to leave the system in the discharge.

System Description

The performance of a particular unit operation of an aquaculture system is often determined by the behavior of the overall system. For example, if an experimental size fine media FBB is operated in parallel with a large rotating biological contactor (RBC) on an aquaculture system, it is possible to have the FBB overloaded with near zero oxygen in the discharge, then wonder why the FBB seems to sludge up and quit working (Miller,1985). The normal substrate concentration in a RBC system which only removes a fraction of the substrate per pass thru the filter would have overload concentrations for a fine media FBB resulting in anerobic conditions in the FBB.

The experimental system consists of two FBB's ( 1.5 m diameter X 2.2 m high) operating at 328 l/min each. The media in these filters is # 90 silica sand which has about 75% bed expansion at 0.3 cm/sec plug flow vertical velocity.

These filters are part of a "feeder guppy" growout system which produces about 15 million fish per year. Along with producing the saleable fish, the system also produces new broodstock ( the poor feed conversion ratios on maturing fish results in a higher percentage of the nitrogen input in the feed showing up in the water rather than as protein in the animal). The water from the FBB's is split into 2 flows, one of which is passed thru a packed column with air for oxygen addition and carbon dioxide removal. The pH of the water increases about 0.25 pH units across the column. Another part of the stream goes through pure oxygen columns to produce high oxygen water. These two streams can enter the culture tanks and are varied according to the load on each individual tank. From the culture tanks, the water flows by gravity to settling tanks for suspended solids removal. Pumps take the water from the clarifiers and return the water to the filter.

The reader should be aware of the fact that the clarifiers are undersized and the suspended solids removal can be considered the weakest link in the system. This weakness effects the performance of the FBB by increasing the solids input which can add to the sludge level in the filter. This weakness also adds to the oxygen consumption of the filter by creating more soluble organics from the decomposition of the suspended solids in the clarifiers. Hence, any results presented on the amount of feed input into the system relative to the filter size and flow rate would not be valid for another system where the suspended solids removal is excellent and no solubilization of the solid waste occurs.

The FBB's on this system introduce the flow to the bottom of the tank via a series of vertical pipes connected to a manifold on the top of the filter. Each one of the vertical pipes can be removed independently of the others and can be changed or maintained without shutting down the filter. This design concept of having the distribution manifold on the top of the tank, rather than on the bottom, is more expensive, but is easier to maintain without shutting down the system. This type of design tradeoff is related to the system in question. In this case, where decades of continuous production is required without the possibility of shutdown, the extra cost is justified.

Since the experiments necessary to test the filters would create conditions which would stress the fish if the test conditions applied to the system as a whole, it was necessary to test only one filter and conduct only short term dynamic tests. For example, to test the pH response of the filter, the pH of the input water to the filter was varied between 5.4 and 8.8. The duration of these tests was kept short enough to allow the fish to stay between 6.2 and 7.8. Considering the value of the livestock in the system, the author was not willing to push the animals beyond this range.

Experimental Design

The first objective in attempting to describe the performance of the FBB is to describe the normal operation. Since the input of nutrients to the filter and the filters metabolic demand vary over a 24 hr period, the response of the filter has been measured over a 24 hr period. The oxygen in, oxygen out and the associated metabolism rate of the filter were continuously measured. Other variables such as ammonia, nitrite, pH, alkalinity were periodically measured throughout the 24 hr period.

The second objective was to measure the ammonia oxidation as a function of pH. Most biological filters demonstrate a well defined pH range within which ammonia is nitrified to nitrate(Kruner, 1983; Sharma, 1977). With a zero net growth biological system with decades of continuous operation, one would theoretically expect the system to have a greater dynamic range than a newly setup and very young, highly loaded filter where surface area competition determines the fitness of the bacteria.

While varying the pH, the metabolism of the filter, pH, ammonia, nitrite, and oxygen were measured at the input and output of the filter.

To properly describe the behavior of a FBB, it was necessary to measure the response to increases in ammonia loading. Exploratory experiments indicated that this filter produced effluent with non-detectable ammonia levels ( less than 0.1 ppm TAN) as long as there was oxygen in the discharge ( greater than 1 ppm). Therefore, it was decided to increase the input oxygen as much as feasible without getting too many gas bubbles from supersaturation and increase the ammonia until we started to see some ammonia in the discharge. Due to mechanical and other limits, I was not able to maintain the high ammonia input for more than a few hours, hence the results of these experiments demonstrate the pulse loading capability of the filter rather than the steady state maximum ammonia oxidation rate.

Another viewpoint for looking at the filter would be to view the behavior as a function of the discharge oxygen levels. Previous experience has shown that this filter produces no ammonia with a zero oxygen point at the top of the sand layer. The experimental objective was to move the zero oxygen point downward in the filter to the point where the system started to produce nitrite or ammonia.

Equipment, Materials and Methods

Oxygen and temperature measurements were made with Royce™ Instruments 9010 and 9040 oxygen meters. Probes from these meters were placed in inlet and discharge streams. The oxygen meters were connected to the serial RS485 communication buss which runs thru the facility. Royce™ portable meters were used to measure verticle oxygen profiles within the filer under low oxygen conditions. Air calibration as per instructions was utilized.

The pH was measured by several different instruments which include Omega™ PHTX-91's connected to Opto 22™ Optomux Brain Board thru a Module AD3. A Cole Parmer series 7142 pH controller and a Jenco model 6009 portable pH meter were also utilized. All pH meters used Innovative Sensors™ 1PB probes. Calibration used standard buffer solutions.

Ammonia and nitrite were measured using Hach™ kits and Sea Test™ test kits for low range and Hach™ nesslers reagent test kit for higher ranges (dilution and the low range kits were also used for the high ranges to check consistency).

Since there is no oxygen input into the filter other than from the input water, the metabolism of the filter can be determined from the oxygen mass balance. This was accomplished via Life Supports™ software, which provided an online real time metabolic output in strip chart format. All data available on the RS485 buss was also collected, graphed, alarmed and archived from the same software.

The RS 485 buss connecting the instruments in the hatchery was connected to an Apple Macintosh™ SE/30 computer via the RS422 serial port on the Mac. The Life Supports aquaculture control system software was responsible for the data collection. This software consists of a collection of objects such as alarms, oxygen devices, temperature devices, pH devices, oxygen valves, feeders, metabolism devices, etc. where each object is relatively independent of the other. Icons, which can be easily moved, associated with the objects are located on a schematic or layout drawings of the hatchery. Activating the icon will display the device, which can be turned on or off, modified, created, or destroyed. Most of the objects can accomplish a wide variety of tasks such as data logging, alarms, verbal alarms with information over the PA system, archiving data, integrating the data (ie feed amounts), handling temperature adjustments to data or feed amounts, etc.

The same software runs all the feeders, controls the oxygen and pH levels levels along with the above monitoring functions which can be used for experimental purposes. However, the normal work load on the computer is presently near maximum capacity ( without either new hardware or major software changes) and there was not enough instruments or computer cycles available to test all the filters and parts of the system simultaneously. With the present high biological load on the system, the risk and cost of devoting more resources to this project was not acceptable.

Results and Discussion

Normal operation:

The normal operation of the system consists of feeding approximately 15 kg/day of SilverCup™ starter and #1 feeds into the system over about 13 hr period. How the waste products from this feed get distributed between the two filters varies depending upon how the tanks are harvested and restocked. Effectively, the two filters get the majority of their input water from separate tanks. This results in a range of input ammonia (N) concentrations from 0.3 to 1.2 ppm. The discharge ammonia concentration is normally non-dectable with the test kits used ( less than 0.1 ppm). The only time that measurable ammonia is detected in the discharge is when there is no oxygen in the discharge. Using one of the portable meters, it was determined that the zero oxygen point in the filter will move down to within 0.8 meters of the bottom before 0.1 ppm ammonia and 0.1 ppm nitrite was detected in the discharge.

When the zero oxygen point in the filter is near the bottom, some nitrite can be measured at the 0.1 ppm (N) level. For various mechanical reasons, it was not possible to sample, measure or rationally analyze the fluid phase below 0.8 meters from the bottom. The near bottom conditions in this filter appear to have some large scale mixing of the input water, thereby giving very erratic measurement. It would be expected that stable operation, with plug flow on the liquid phase would take a distance to establish that would be proportional to the distance between water injection points ( about 0.5 meters).

Because the FBB's are followed by packed columns, it is possible to get an indication of the general BOD removal of the filter by observing the biofouling on the packing material. As long as the discharge oxygen from the filter is under feedback control with a set point of 0.5 ppm oxygen, there is no indication of any biofouling or large scale biomass growth in the return piping system. Measurments of the mass transport coefficients associated with the packed columns indicate an alpha of 1.0 ( ie, the discharge water behaves like pure water), whereas the FBB input water will have an alpha in the 0.5 to 0.75 range when the system is being fed.

In terms of normal operation procedures, ammonia in the discharge is not monitored, expect when the oxygen in the discharge goes to zero. These operating procedures have produced satisfactory results without a system crash.

Input pH variation:

The pH of the input water was increased by adding soda ash and decreased by adding HCl, via a variable speed chemical feed pump, to the input water. The resulting pH was measured and the feed rate of the pump adjusted until the desired pH was achieved. The range of pH tested was between 8.8 and 5.4 on the inlet water. During these experiments the input ammonia (N) was between 0.8 and 0.9 ppm and the discharge oxygen was maintained above 3.7 ppm ( 3.7 to 6.3 ppm in the discharge). The temperature was 25 centigrade and the starting alkalinity was 2 meq/l with a starting ph of 6.8. The input and output ammonia, nitrate, and oxygen were measured and the metabolism of filter was calculated.

The results are very unexciting. No ammonia was detected in the discharge while increasing the pH. Decreasing the pH finally did show a fall off in nitrification at a pH of 5.35 where the discharge increased to 0.5 ppm with a 0.9 ppm inlet. Once breakthru was achieved, soda ash was pumped into the system in order to return the system to normal. Guppies will tolerate this low pH water but they are stressed. Upon increasing the input pH, the ammonia returned to non-detectable levels.

The decrease in ammonia oxidation is reflected in the decrease in metabolism rate of the filter. This effect is shown in Figure 1.

Figure 1. Screen Dump of Metabolism of the filter as the pH was decreased. The metabolism is in gm/hr of oxygen consumption.

Ammonia Loading:

One way to measure the performance of a biological filter is to increase the loading and monitor the response. This approach was accomplished by adding ammonium sulfate to the input water to the filter with a chemical feed pump. As previously mentioned, this filter normally produces non-detectable ammonia concentrations in the effluent as long as there is oxygen in the discharge. To obtain reasonable results with oxygen in the discharge it was necessary to add some 200 ppm oxygen water to the input.

This series of experiments was conducted at a temperature of 25 centigrade with an input pH of 6.84 with 2.0 meq/l alkalinity. The input oxygen levels were increased to the 18 to 20 ppm range. These very high supersaturation levels of oxygen did create gas bubbles in the FBB, so the true metabolism rate is not fully known. The results are shown in Table 1.

Table 1 Ammonia Loading Experiments

The results are confusing in that no ammonia was detected at 4.2 ppm input while there appeared to be 0.1 ppm at 3.6 ppm ammonia. This behavior may indicate the possibility of luxury consumption on the part of the nitrification bacteria. However, the main factor to observe is that the peak rate of nitrification is about 4 times the average rate. This observation is consistent with the zero net growth concept previously discussed.

It is believed that the bacteria in this filter are literally starving with little or no net production of biomass. When presented with extra nutrients, it is easy for the existing bacteria to increase their consumption by a factor of 4 for a short time.

The behavior of this class of filters toward pulse loads can be very useful for certain classes of business such as temporary holding facilities, depuration facilities and similar operations where the loading varies over very short time periods. An example of such a situation is a goldfish wholesale operation. Such a facility will receive large shipments on Monday and Wednesday and be sold out by Friday. Fine sand FBB's at two such facilities have demonstrated shock load capability without crashing. One of the facilities had a human error which had a full tank of fish shut off from the system and died. The fish rotted for three days and then someone put the tank back on the recycle system. Within two hrs. the entire system was free of ammonia in the filter discharge water.

Other Experience:

Similar FBB's are operated by other people with similar results. The Spring and Groundwater Institute in Shepherdstown, WV. operates two filters, the same size at the filter tested in this study, at 740 l/min each (larger media). A typical data set shows input TAN of 0.7 ppm with an output of 0.04 to 0.06 ppm. This system is being feed 35 lb/day of feed and has microscreening suspended solids control. This means that the FBB's see primarily ammonia and very little general BOD. This is reflected in the delta oxygen across the filter in the 3.3 ppm range.

Installations on fish holding systems have shown exceptional stability in the face of highly variable loads. Two local installations were tested by the author and both had < 0.1 ppm TAN in the effluent with inlet values between 0 and 1.2 ppm. These are unfed systems and suspended solids removal is not a major problem.

Another unit on a tilapia broodstock facility has demonstrated excellent nitrite removal capability. With the large volume and internal structure in this facility, most of the ammonia is converted to nitrite in the system, which in tujrn became the limiting factor. Installation of a 350 l/min FBB on the system eliminated this nitrite problem.

Some very highly loaded systems have experienced some stability and performance problems that are difficult to understand except in the context of very high loading, poor suspended solids removal (slip stream SS removal rather than full flow) in the balance of the system and a physical design that is hydraulically different than the systems described above. This system is running deeper filters at higher velocities with a less well graded media.

Conclusions

Fine media fluidized bed biological filters have demonstrated the highest level of effluent water quality and the greatest dynamic range to pulse loading of any of the biological filters used in aquaculture. Whenever high quality water is desired, fine media fluidized bed biological filters should be part of the system.

References