Study of the Correlation of Mixed Liquor Volatile Suspended Solids (MLVSS) and Turbidity

Study of the Correlation of Mixed Liquor Volatile Suspended Solids (MLVSS) and Turbidity

Study of the Correlation of Mixed Liquor Volatile Suspended Solids (MLVSS) and Turbidity

Grove and Skopp Associates, Inc.

Amy Grove, Associate

Kendra Skopp, Associate

Project Deadline: 5/12/2003

Abstract

A model wastewater treatment plant was assembled as a sequencing batch reactor that would clean synthetic waste using activated sludge. The entire plant was controlled by a software program written in LabVIEW. The NRP Software ran continuously on a computer. A Honeywell turbidity sensor was connected to the software. The objective was to find a functional relationship between the voltage output from the turbidity sensor and Mixed Liquor Volatile Suspended Solids (MLVSS) concentration. In order to obtain a relationship between turbidity and MLVSS, a range of turbidity readings for various MLVSS levels was needed. A unique voltage could not be found for each MLVSS level. A 1 cm3 acrylic piece was placed in between the laser and the detector on the sensor to allow fewer solids to diffract the laser light. Again, no unique relationship was found. During settle phase, the data became more variable after the acrylic piece was added. Sometimes it would sharply increase in voltage, while other times it decreased. The aeration voltage increased by approximately two volts. The manner in which the solids would settle was not the same, sometimes the solids would settle through the sensor and other times not. The sensor was very vulnerable to biofilms forming over the detector. No functional relationship was found between MLVSS and the turbidity. The closest model was exponential with a R2 value of 0.5667. Therefore, other sensor options were explored. Further research should be done with these sensors.

Introduction

This is a research project done for CEE 453, Laboratory Research in Environmental Engineering. A model wastewater treatment plant was set up in the laboratory that would clean synthetic waste using activated sludge from the Ithaca Wastewater Treatment Plant. With this model, we were attempting to find a correlation the Mixed Liquor Volatile Suspended Solids (MLVSS) in the reactor and the turbidity. To do this we utilized a Honeywell Turbidity Sensor. We then hoped to incorporate our findings into the Main NRP software used to control the system.

Objective

Our original objective was to find a relationship between the voltage output from the Honeywell turbidity sensor and MLVSS concentration in the reactor. Our hypothesis was that there was a functional relationship between MLVSS and turbidity, most likely linear or exponential. Once this relationship was found, we wanted to incorporate a turbidity measurement into our NRP software and optimize the amount of MLVSS in our reactor. To optimize the MLVSS, we would use the turbidity measurement to add or remove activated sludge. However, during the course of our research, our objectives changed slightly. It became apparent that before we could continue with our original research we needed to find a practical method to keep the sensor in the reactor and obtain accurate and consistent data.

Materials and Methods

The Reactor

The reactor used for this research was modeled after a Sequencing Batch Reactor. It consisted of a 6L square tank with approximately 2L of activated sludge and 2L of 1X synthetic waste. The waste consisted of:

0.0844 g/L / Starch / 0.07533 g/L / Ammonium chloride
0.125 g/L / Casein / 0.0069 g/L / Potassium phosphate
0.0319 g/L / Sodium acetate / 0.175 g/L / Sodium hydroxide
0.0116 g/L / Capric acid / 0.012 g/L / Glycerol

as specified in the Laboratory Research in Environmental Engineering Laboratory Manual.

We used a Cole Parmer model peristaltic pump to deliver the waste in a 20X concentration to the reactor. It was further diluted to 1X in the reactor by pumping in tap water with the peristaltic pump.

The aeration system consisted of an air supply, solenoid and needle valves to control the flow rate, an accumulator, pressure sensors to measure the pressure immediately upstream from the diffuser stone and the pressure at the effluent of the accumulator. The reactor was connected to the air supply system. In the reactor a dissolved oxygen and temperature probe were submerged. The apparatus was set up according to Figure 1 below.

Figure 1. Apparatus used to aerate reactor and measure dissolved oxygen.

From Laboratory Research in Environmental Engineering Laboratory Manual 7th Ed.

The drainage system for the reactor consisted of a small pipe inside the tank. It was set at the correct angle so it would drain out only the effluent after settling and not drain the activated sludge. It worked by gravity only; no pumps were used in the drainage system.

The Software

The entire plant was controlled by a software program written in LabVIEW. The NRP Software ran continuously on a computer. There were six states the plant could be operating in when in automatic operation. The plant could also be operated in an operator selected state where the plant states could be manually changed. The first stage was for the plant to fill with waste until the pressure sensor in the bottom of the tank measured an equivalent voltage to 2100 mL. The second stage was for the plant to fill with tap water until the volume was 4000 mL. The configuration took into account extra volume added by sensors in the reactor. The reactor then went into aeration mode. It aerated for 20000 seconds or 5.5 hours. The amount of aeration could be controlled by setting the air flow rate (moles/sec) or by the dissolved oxygen in the tank measured by the Dissolved Oxygen probe. After the aeration state, the tank settled for 3600 seconds or 1 hour. The tank then drained the effluent off the top of the activate sludge until it reached a volume of 2000 mL. The plant began the cycle again in the fill with waste setting. For the settings not primarily controlled by time, there was a maximum amount of time that particular state could be running. This was used as a back up in case of any failure of the setup or sensors.

The Sensor

For our research, we placed a Honeywell Turbidity Sensor in the reactor. The sensor sends an infrared (Class 3b) laser across a 2.4 cm space. Straight across from the laser is a light well that collects the beam. At a ~10° angle, there is a detector that collects and registers any light that had been diffracted. For the configuration set up of our reactor, there is a volt range of 0-5. As turbidity increases, the voltage increases until there is a point where there are too many solids and it blocks some of the light. The voltage then decreases. Initially the sensors electronics were exposed and it needed to be waterproofed. Lee Virtue of the Civil Engineering Shop enclosed our sensor in a waterproof case. The first sensor tested by us began to leak, but after some minor repairs to the design, it was fully functional. Figure 2 below shows the front and back of the sensor used.

Figure 2. Honeywell Turbidity Sensor as used in the reactor.

The sensor is very sensitive to light so the reactor was boxed in, so that the light allowed to the sensor would be more consistent throughout the day.

The Dilution Process

In order to obtain a relationship between turbidity and MLVSS we needed to get a range of turbidity readings for various MLVSS levels. We started by taking a measurement of the normal reactor contents at a full strength. We diluted the contents by a factor of 2 in a separate tank and took readings on each dilution. This process was repeated 5 times until we had a 1:32 dilution. MLVSS measurements were taken at the same time by conventional methods as defined by the APHA; AWWA; WEF (2000). The protocol we used is defined below:

  1. Using filter tweezers, transfer a prewashed and dried filter (Whatman GF/or equivalent) to balance and determine tare weight to nearest 0.01 mg.
  2. Place in a Millipore® filtration apparatus, and turn on vacuum supply.
  3. Pour through a 20 mL sample.
  4. Wash sample container, then the filter holder and filter, with two 10 mL samples of distilled water.
  5. Carefully remove filter with tweezers when finished.
  6. Dry for 1 hour at 105°C using an aluminum dish to hold the filter in drying oven.
  7. Reweigh to nearest 0.01 mg.
  8. Compute TSS as g/L (mass residue/volume of initial sample).
  9. Dry for 20 minutes at 550°.
  10. Compute MLVSS as g/L (mass residue/volume of initial sample).

During dilution, all of the solids were kept in separate storage tanks. When dilution was done, all storage tanks and reactor settled for an hour, and then the solids were replaced back into the reactor.

Results and Discussion

The first dilution process was done with the reactor to see what the sensor voltage curve would produce. Figure 3 below shows the results of that dilution.

As can be seen from this figure, each voltage does not correspond to a unique dilution. This is a problem since we wanted to use the voltage reading to correlate with a unique MLVSS measurement. We hypothesized that the rector contents were so dense after a certain point that instead of increasing voltage and diffracting the laser, the laser was actually blocked causing a voltage decrease.

The turbidity sensor was very insensitive at low dilutions. This can be seen by the first three data points. The graph is nice and smoothly rounded and then suddenly flat lines. This is not good because one of our potential applications of the sensor is to determine when the reactor has settled past a certain point. However, if the sensor cannot determine between a 1/16th dilution and clear water it cannot be used for this purpose.

To hopefully solve these problems, a 1 cm3 acrylic piece was placed in between the laser and the detector on the sensor. This piece would allow fewer solids to diffract the laser light, thus keeping the voltage to the left side of the above curve and creating a unique voltage for every dilution. Once the acrylic piece was in place, the dilution process was redone. The new results are shown in Figure 4 below.

Now, the sensor is more sensitive at the smaller dilutions. It also does not go over the peak point of the voltage. This means that each dilution has a unique turbidity reading.

However, at a later dilution process, a problem was realized. As the MLVSS grew in the reactor, the curve once again went over the peak point. This is shown in Figure 5. The acrylic piece worked up to a point, however it is necessary to have a voltage that always has a unique reading. Otherwise, the NRP software could not be set up to use the voltage in a practical manner. The acrylic piece does solve the problem of low sensitivity with the smaller dilutions. However, the MLVSS would need to be kept below the peak point at all times to use this set-up.

We also encountered other sensor problems during the course of our research. The sensor face was extremely vulnerable to biofilms and waste build-up. After a few days of the sensor housing sitting in the reactor, the lightwell and detector would become covered with sludge. This caused the data reading to be inaccurate since it always appeared there was a high turbidity even if there was not. Figure 6 below shows pictures of the sensor after 1 day and after 5 days of sitting in the reactor without being cleaned.

Figure 6. Various sensor views after being placed in reactor without cleaning.

As can be seen on the side view after 5 days, there is a considerable amount of build-up between the laser and the detector. The sensor would have to be cleaned at least every several days, which would create a lot of maintenance for a plant operator, especially if the sensor was in a large reactor and submerged at a significant depth.

During the course of a day (24 hours), data from the turbidity sensor was logged approximately every minute. Data was taken in all stages of the reactor. When the sensor voltage was plotted against time several things were apparent. Figure 7 below shows the data logged when there was no acrylic piece in the sensor and when it had been cleaned.

In this graph, the settle period is clearly defined with a sharp increase in voltage from the turbidity sensor. Even though the actual voltages vary among the three, they are relatively much higher than the base line voltage during the aeration phase. When the acrylic piece was added there was much more noise in the data as can been seen in Figure 8.

During the aeration states the voltage varies within a 0.75 volt range. With no acrylic piece, the voltage only varied in a 0.2 volt range. The aeration voltage average increased from approximately 1.4 volts to 3.6 volts. This increase was because we were moving from the right side of the curve in Figure 3 towards the peak point.

During the settle phase, the voltage output was very irregular. Sometimes it would sharply increase in voltage, while other times it decreased. This could be due to a number of factors. It could be caused by problems introduced by the acrylic piece. The light going through the acrylic piece may not have traveled in a straight line, but rather been diffracted in any direction causing more variability. Also during the settle phase, it was noticed that solids would not always settle through the turbidity sensor, but settle on top. Since the solids often flocculate on top of the sensor, it could have blocked the detector causing a false high output. Also it could have inhibited more solids from passing through the sensor creating a pocket of more clear water in the sensor and thus causing a false low output. Figure 9 illustrates this point. The manner in which the solids would settle each time was not always the same, sometimes the solids would settle through and other times they would not.

Another possibility to explain the variability in sensor voltage could be due to the biofilms mentioned earlier. Due to all the variability and non-uniqueness of voltage experienced with the acrylic piece, the turbidity sensor as is does not appear to be a viable long term option for measuring turbidity or MLVSS.

Despite our conclusion that this is not a viable option for a wastewater treatment plant, we still tried to find a functional relationship between the MLVSS and the turbidity output that we recorded. With the useable MLVSS concentrations measured from various dilutions and the averages of the voltage recorded, Figure 10 was obtained. We used only points that we knew were on the left side of the curve of Figure 3.

Using Microsoft Excel's trend line feature, we attempted to fit several models to the curve. There was no clear functional relationship found. Although the graph looks exponential, the R2 coefficient is equal to 0.5667. Both the linear model and the power model gave even worse results.

Due to the deficiencies of this sensor, other sensors options were researched online for potential further exploration. If given more time a sensor to test would be the Forest Technology Systems DTS-12 sensor. This has a self cleaning face and is a reasonable size. It is designed to run for long periods of time and can withstand depths of up to 98 feet, well beyond what is needed for this application. The TROLL 9000 from EnviroEquip would also be a good model to look into. It has already been tested and proven to give accurate and reliable data. The size is reasonable and can also monitor things such as temperature, pressure, depth, dissolved oxygen, conductivity, pH, and other various parameters. A third option would be from Intermountain Environmental, Inc., model WQ710. It is a submersible sensor that also has an automatic lens cleaner. It uses the Ratiometric method which takes into account ambient light and virtually eliminates this source of error, and therefore the reactor may no longer have to be black boxed if this sensor was used. The Ratiometric method uses two detectors and records the ratio; therefore, it is independent of laser intensity and other light.

Conclusion

Our research project started out with the goal of finding a useable functional relationship between MLVSS and turbidity, but turned into a feasibility study of the Honeywell turbidity sensor in an actual wastewater treatment plant. We began our research by doing the successive dilutions of the solids in the reactor. From those tests, we decided to make the length of which the solids could pass through the sensor smaller. This was to create a unique voltage for each MLVSS level. We were able to create a unique curve, but only up until a certain point. The turbidity of the reactor grew each day because MLVSS increases over time, and soon was again too high for the sensor.

In the process of research, it was found that when the sensor sat in the reactor for multiple days without cleaning or scraping, a biofilm formed over the acrylic piece and the detector. By looking at the turbidity voltages recorded over a day, it was concluded that the sensor would only be viable if cleaned everyday, a time consuming operation for a plant operator. Even when clean, the sensor still had large variability during the settle phase with the acrylic piece. Overall, the sensor did not turn out to be a very accurate or consistent measure of MLVSS in the reactor. It consistently gave unreliable data for several reasons.