Citation
The Effect of Flow Rate on Silt Density Index, and Correlations between Silt Density Index, Particle Count, and Conductivity

Material Information

Title:
The Effect of Flow Rate on Silt Density Index, and Correlations between Silt Density Index, Particle Count, and Conductivity
Creator:
Hodgkins, Daniel Ross
Publication Date:
Language:
English

Notes

Abstract:
Water filtration via reverse osmosis (RO) filters is essential in a multitude of industrial processes where extremely pure water is required. Water quality at the RO inlets is closely monitored to ensure that RO units can operate effectively. The most common water quality measurement taken to do so is the Silt Density Index (SDI), which is a measurement of the potential for RO membrane fouling. If SDI becomes too summa (>5), normal operation must be temporarily stopped until SDI lowers. Therefore, it is imperative that any system variables affecting SDI be identified. Additionally, since SDI cannot be monitored in live time, finding correlations with online analyzers can be useful when predicting SDI values. In this thesis, the effects of doubling system flow rate on SDI are explored first. After performing a hypothesis test, it can be confidently stated that doubling system flow rate increases SDI in a statistically significant way. Additionally, correlations between SDI, particle count, and conductivity are analyzed. Particle count has a weak positive correlation to SDI and correlation strength increases as analyzer location distance to the RO inlets decreases. Conductivity has a weak to moderate positive correlation to SDI. ( en )
General Note:
Awarded Bachelor of Science in Chemical Engineering, summa cum laude, on May 8, 2018. Major: Chemical Engineering
General Note:
College or School: College of Engineering
General Note:
Advisor: Sergey Vasenkov. Advisor Department or School: Chemical Engineering

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Daniel Ross Hodgkins. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

UFDC Membership

Aggregations:
UF Undergraduate Honors Theses

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

The Effect of Flow Rate on Silt Density Index, and Correlations between Silt Density Index, Particle Count, and Conductivity By Daniel Hodgkins

PAGE 2

1. Abstract Water filtration via reverse osmosis (RO) filters is essential in a multitude of industrial processes where extremely pure water is required. Water quality at the RO inlets is closely monitored to ensure that RO units can operate effectively. The most common wa ter quality measurement taken to do so is the Silt Density Index (SDI), which is a measurement of the potential for RO membrane fouling. If SDI becomes too high (>5), normal operation must be temporarily stopped until SDI lowers. Therefore, it is imperativ e that any system variables a ffecting SDI be identified. Additionally, since SDI cannot be monitored in live time, finding correlations with online analyzers can be useful when predicting SDI values. In this thesis, the effects of doubling system flow rate on SDI are explored first. After performing a hypothesis test, it can be confidently stated that doubling system flow rate increases SDI in a statistically significant way. Additionally, correlations between SDI, particle count, and conductivity are analy zed. Particle count has a weak positive correlation to SDI and correlation strength increases as analyzer location distance to the RO inlets decreases. Conductivity has a weak to moderate positive correlation to SDI.

PAGE 3

2 Introduction 2 .1 Background Information Water filtration the removal of impurities from water, is an essential aspect of numerous industrial processes including desalination, waste water treatment an d power production. Specifically pow er plants that utilize steam initially filter process water in order to avoid corrosion of boiler tubes and turbine blades. For processes that require extremely pure water, reverse osmosis (RO) filters are typically imple mented as one of the last filtration steps RO filter s apply a pressure to the incoming through a semi permeable membrane of pore size ~0.001 microns [1] Therefore, only particles less than 0.001 microns in size remain in solution after the filtration process has been completed. Because o f the membrane fouling (the deposition of particles on a membrane) can occur rapidly if the RO inlet water quality is poor. If fouling occurs too quickly, RO units will not properly filter process water, leading to damage/corros ion of the system downstream. Membranes also need to be replaced more frequently if the rate of fouling is high. To prevent fouling, pre filtration units are typically installed upstream of the RO units to remove larger particles in the water before reachi ng the RO units. Examples of these filters include multi media filters (MMFs), sand filters (SFs), decarbonators, and cartridge filters (CFs). Coagulant injection upstream of the pre filters is common procedure as well, which clumps smaller particles in so lution together to form larger particles. These larger particles can then be filtered out by the pre filters instead of reaching the RO units. Even with coagulant injection and pre filte r units, smaller particles that make it to the RO units may still lea d to rapi d membrane fouling To ensure the proper precautions are taken when water quality is not sufficient, daily monitoring of RO inlet water must be performed. Measurement of the s ilt density index (SDI) is an industry standard for monitoring RO inlet water quality, and indicates the potential for rapid membrane fouling. To take an SDI measurement, a bleed off line of RO inlet water is pumped to 30 psig and run through a 0.45 micron filter [2] The time required for a pre determined volume of water, V, to pass through the filter at the start of the procedure is recorded as t i [2] After 1 5 minutes of operation (T=15) the time required for the same volume of water to pass through the filter is recorded as t 1 5 [2] The time required should be larger at 15 minutes of operation since particles have had more time to deposit on the filter, impeding water to pass through [2] Note that s ometimes SDI is calculated based on other elapsed operation times but T = 15 min is common practice [2] SDI is given b y Equation 1 [2] (1)

PAGE 4

If an SDI measurement is greater than 5, significant membrane fouling is said to occur and filtration plants either temporarily stop filtration or switch to an alternative water source such as city water until the SDI goes below 5 again. Temporary shutdown is often impractical when a constant supply of filtered water is needed and city water is much more expensive than primary sou rces such as reclaim or sea water. Therefore, exploring whether certain system variables have an effect on SDI is beneficial so that these variables can be adjusted accordingl y in order to help lower SDI when it is too high for normal operation. Addition ally, b ecause of the nature of the measurement procedure, SDI online analyzers do not exist and SDI measurements are generally only taken a handful of times a day. SDI may fluctuate between these measurements, reaching values higher than 5 and potentially damaging the system. However, other online analyzer values such as particle count or conductivity may serve as an indication of high SDI. For this reason, any correlations bet ween analyzer values and SDI are valuable to be aware of. 2 .2 Scope of Thesis The effects of system variables on SDI and SDI correlations to certain online analyzers are explored in this thesis Specifically, a statistical analysis is performed on whether having 1 RO unit online versus 2 (i.e. doubling volumetric flow rate) significant ly affects SDI Furthermore, c orrelations between SDI and particle count online analyzer readings are looked into. Correlations between SDI and conductivity analyzer reading are investigated as well. Data used in all analyses was collected from the UF Coge 2 .3 UF Cogeneration Plant Filtration System The UF Cogeneration Plant (Cogen) produces power from a 50 MW combustion turbine and also provides a constant supply of steam to the UF campus. The water used to make the st eam must be extremely pure to prevent system corrosion and therefore must first be filtered by mu ltiple filtration units onsite. At the inlet of the filtration process, free reclaimed water is supplied to the Cogen from the UF Wastewater Treatment Plant ( WWTP). Coagulant injection occurs at th e Cogen inlet and the water proceeds to go through multi media filters (MMFs) sand filters (SFs) a decarbonator (Decarb) cartridge filters (CFs) and RO filters. A nominal flow rate of approximately 135 gallons per minute (GPM) per RO unit is set by flow control valves on each RO filter Therefore, switching from 1 RO unit online to 2 doubles the total system flow rate from 135 GPM to 270 GPM. After the RO filters, water is sent to a storage tank which controls whet her 1 or 2 RO filters are online based on 2 level transmitters within the tank When 1 RO unit is online, storage tank level gradually decreases until a lower bound level is reached which signals a 2 nd RO to turn on. Similarly, when 2 RO units are online, storage tank level gradually increases until an upper bound level is reached which signals the 2 nd RO to be turned off. There are 3 RO units total (RO A, B, and C), but a maximum of 2 are on at once. Figure 1 displays a simplified process flow diagram of t he water filtration system.

PAGE 5

Figure 1: Process Flow Diagram of the Cogen Filtration System 3 Statistical Analysis and Discussion SDI data and time s of measurement are logged by Cogen operators daily. This data was extracted between July, 2016 and March, 2017 for analysis resulting i n a total SDI sample size of N = 169 Online analyzer data was extracted from Plant Information (Osisoft software) at t he times the SDI measurements were taken as well. Analyzers that displayed whether each RO unit was online were used to determine the number of ROs online at each SDI measurement. Particle count and conductivity analyzer data were collected for correlation analysis. All statistica l analysis was performed via SAS software. 3 .1 SDI Frequency Distribution T he frequency distribution of all SDI data collected was graphed and = 3.9 and = 0.95. Both the distribution and normal curve are displayed in Figure 2. Coagulant MMFs SFs s Decarb CFs Storage Tank Reclaim RO A RO B RO C

PAGE 6

Figure 2: Frequency Distribution and Normal Curve of Collected SDI 3 .2 Effect of Number of ROs Online on SDI 3 .2.1 Statistical Analysis SDI data was first sorted based on whether 1 or 2 RO units were online at the time of SDI measurement. This data categorization resulted in a sample size of N = 71 for 1 RO units online and N = 9 8 for 2 RO units online. Frequency distributions were then graphed for both scenarios along with fitted normal curves These distributions, normal curves values are shown in Figures 3 and 4.

PAGE 7

Figure 3: SDI Frequency Distribution and Normal Curve for 1 RO Unit Online Figure 4 : SDI Frequency Distribution and Normal Curve for 2 RO Units Online

PAGE 8

Given the fairly different distributions and values, a one tailed null hypothesis (H 0 ) and alternative hypothesis (H 1 ) were formed: H 0 : 1 RO 2 RO H 1 : 1 RO 2 RO < 0 In other words, the null hypothesis states that the population mean for 1 RO SDI values is greater than or equal to the population mean for 2 RO SDI values. On the other hand, the alternative hypothesis claims that the population mean for 1 RO SDI values i s less than the population mean for 2 RO SDI values. To decide whether to retain H 0 a t test was performed. Note that this test assumes that both populations, 1 RO and 2 RO SDI values are normally distributed. Table 1 lists general statistical results a nd Table 2 shows t and p values for equal and unequal population variances. Both scenarios le d to a p value of <0.0001 which is equivalent to saying that the chances of obtaining the collected SDI values under the null hypothesis are less than 0.01%. There fore, the null hypothesis can confidently be rejected and it can be said that the number of ROs online has a statistically significant impact on SDI values. ROs Online Method N Mean Std Dev Std Err Minimum Maximum 1 71 3.5418 0.9614 0.1141 1.1100 6.1500 2 98 4.1558 0.8528 0.0861 2.5200 5.8800 Diff (1 2) Pooled 0.6140 0.8999 0.1402 Diff (1 2) Satterthwaite 0.6140 0.1430 Table 1: General Statistical Results for 1 RO and 2 RO SDI Data Method Variances DF t Value P Value Pooled Equal 167 4.38 <.0001 Satterthwaite Unequal 139.78 4.29 <.0001 Table 2: Hypothesis t Test Results

PAGE 9

3 .2.2 Discussion Statistical analysis led to the statement that having 2 RO units online versus 1 almost certainly increases SDI values. As previously mentioned, the underlying variable in this case is volumetric flow rate, which is approximately 135 GPM when 1 RO unit is online and doubles to 270 GPM when 2 RO units are online. An increase in SDI when flow rate is doubled could be attributed to a decrease in pre filter system performance. MMFs, SFs, and other filtration units upstream have to filter twice the amount of wa ter in the same duration of time when 2 RO units are online compared to 1. Particle build up in these filters potentially occurs twice as fast, resulting in worse filtration and an increase in particles making it to the RO inlets. A higher concentration of particles at the RO inlet s of course leads to a higher SDI. Another potential reason that SDI is generally higher at a doubled flow rate involves the coagulant injection at the filtration system inlet. At twice the flow rate, twice the amount of coagulant should be injected in order to effectively flocculate (clump together) the doubled flux of particles. Whether this adjustment in coagulant injection actually occurs should be further explored. If SDI becomes too high for normal operation, the Cogen should consider remaining on 1 RO unit for as long as possible to help lower SDI. Of course, the storage tank downstream will eventually become too low from the insufficient flow rate of 1 RO unit. To replenish the storage tank yet operate at a lower flow rate, a 2 nd RO could be turned on with a temporary decrease in flo w rate set point. This set point would be optimized when storage tank level is at steady state i.e. stagnant. 3 .3 SDI Correlations to Particle Count Analyzer Data 3 .3.1 Statistical Analysis Onli ne particle count analyzers are located at the let, MMF outlet, and decarbonator outlet. Each analyzer measures the concentration of particles ranging from 2 125 microns in particles/mL at their respective locations. To determine whether correlations exist between SDI and particle count s analyzer data was extracted at the times of each SDI measurement. Along with the 3 analyzer readings, t he difference between MMF inlet and outlet particle count (Delta MMF) readings was also plotted against SDI in search of a correlation Table 3 displays P earson corr elation coefficients (and confidence limits) for each particle count variable that was correlated to SDI. The MMF inlet particle count showed a very weak positive correlation to SDI while the delta MMF particle count indicated a very weak negative correlat ion. The decarb outlet count showed a stronger positive correlation than the MMF outlet count, but both correlations to SDI were still considered weak

PAGE 10

Variable With Variable N Correlation (r) 95% Confidence Limits MMF Inlet Particle Count SDI 169 0.13728 0.014374 0.282016 MMF Outlet Particle Count SDI 169 0.30651 0.162210 0.436502 Delta MMF Particle Count SDI 169 0.17660 0.318585 0.025820 Decarb Outlet Particle Count SDI 169 0.35263 0.212008 0.477328 Table 3: Pearson Correlation Coefficients for Particle Count Data Figures 5 8 display SDI plotted against each particle count variable. Figure 5 : SDI vs. MMF Inlet Particle Count

PAGE 11

Fig ure 6 : SDI vs. MMF Outlet Particle Count Fig ure 7 : SDI vs. Delta MMF Particle Count

PAGE 12

Fig ure 8 : SDI vs. Decarb Outlet Particle Count Although none of the correlations to SDI were strong, null hypothesis z transformation tests concluded that correlations more than likely exist. For each correlation, the null and alternative hypotheses proposed were: H 0 : r = 0 H 1 In other words, t he null hypothesis stated that no correlation exists while the alternative hypothesis claims that a correlation does exist. Table 4 shows the resulting p values for each hypothesis test. With great confidence, H 0 can be rejected for MMF outlet and decarb o utlet particle count correlations. Therefore, a correlation to SDI almost certainly exists for both of these variables. Choosing a significance level of 0.05, the delta MMF particle count correlation is significant (p<0.05) and a correlation to SDI is said to exist i.e. H 0 is rejected. However, the MMF inlet particle count p value le ads to the null hypothesis being retained with a p value greater than 0.05.

PAGE 13

Variable With Variable H 0 : r = 0 p Value Inlet MMF Particle Count SDI 0.0751 Outlet MMF Particle Count SDI <.0001 Delta MMF Particle Count SDI 0.0215 Outlet Decarb Particle Count SDI <.0001 Table 4: Hypothesis z Test Results for Particle Count Correlations 3 .3.2 Discussion As expected, particle count at each location correlated positively with SDI. Furthermore, the delta MMF particle count correlated negatively with SDI as expected since SDI should decrease with increasing amount of particles removed by the MMFs. Another ob servation is that the particle count correlation beco me s stronger as dis tance to the RO inlets decreases This effect makes intuitive sense because SDI should correlate best with the particle count at the location of the SDI measurement (the RO inlet). On the other hand, a high particle count at the MMF inlet may not correspond to a high SDI because if the majority of the particles at this location happen to be large, they will be filtered before reaching the RO inlets and will therefore not have any effect on SDI. Even at the decarb outlet near the RO inlets, the correlation is considered weak One explanation is that particle count analyzers have a detection limit size of 2 microns. Therefore, if the majority of the particles are below this limit at the R O inlet, a low particle count could still result in a high SDI. For an indication of SDI the C ogen (and other filtration plants) should focus primarily on the decarb outlet particle count because it correlates best with SDI values due to its close location to the RO inlets. Furthermore, i nstallation of another particle count analyzer right at the RO inlets to provide a better corre la tion could be considered However, particle count analyzers in general should not be heavily relied upon when estimating SDI because of their weak correlations to SDI measurements. 3 .4 SDI Correlations to Conductivity Analyzer Data 3 .4.1 Statistical Ana lysis Online conductivity analyzers are located in Cogen RO units A, B, and C, and measure conductivity in SDI data was plotted against RO A conductiv ity only when RO A was online. Similar logic follows for RO B and C data, leading to 3 correlation analyses corresponding to each RO unit. Table 5 shows Pearson correlation coefficients (and confidence limits) for each set of conductivity analyzer readings correlated

PAGE 14

to SDI. RO A and B conductivity display moderate positive correlations to SDI while RO C conductivity shows a weak positive correlation. Variable With Variable N Correlation (r) 95% Confidence Limits RO A Conductivity SDI 92 0.46802 0.288747 0.612374 RO B Conductivity SDI 91 0.41727 0.228992 0.572337 RO C Conductivity SDI 84 0.26785 0.055100 0.454775 Table 5 : Pearson Correlation Coefficients for RO Conductivities Figures 9 11 display SDI plotted against conductivity Figure 9 : SDI vs. RO A Conductivity

PAGE 15

Figure 10 : SDI vs. RO B Conductivity Figure 11 : SDI vs. RO C Conductivity

PAGE 16

The same hypothesis test constructed for particle count correlations was performed on conductivity correlations to SD I Since p values for RO A and B conductivities are less than 0.0001 the null hypothesis t hat no correlation exists can confidently be rejected for both cases. If a significance level of 0.05 is chosen again, the null hypothesis can also be rejected for RO C conductivity correlat ion to SDI (p<0.05). Table 6 lists the resulting p values from the hypothesis test. Variable With Variable H 0 : r = 0 p Value RO A Conductivity SDI <.0001 RO B Conductivity SDI <.0001 RO C Conductivity SDI 0.0135 Table 6: Hypothesis z Test Results for RO Conductivity Correlations 3 .4.2 Discussion RO conductivity had a positive correlation to SDI which was to be expected. Particles in the water have an associated conductivity and an increase in the concentration of particles should of course increase both SDI and conductivity. While the conductivity correlations were generally stronger than that of particle count, they were still only moderate at best. An explanation could be variability in particle composition, which conductivity is heavily dependent on. The Cogen (and other filtration plants) should definitely monitor RO conductivity along with decarb particle count in order to get an idea of current SDI value. However, the conductivity correlatio ns are not strong enough to be heavily relied upon. 4 Conclusion In summation, having 2 RO units online versus 1 i.e. doubling system flow rate, increases SDI val ues in a statistically significant way To combat this effect, filtration plants should operate on 1 RO unit as often as possible and should consider whether a temporary decrease in flow rate set point is plausible when SDI is high. Furthermore, SDI correlates best with the decarb outlet particle count and RO conductivity, and these analyzer readings should be used as a loose indication of current SDI values.

PAGE 17

5. Works Cited [1] Puretec Industrial Water : Ultrapure Water Solutions puretecwater.com/reverse osmosis/what is reverse osmosis. [2] WaterProfessionals www.waterprofessionals.com/why waterprofessionals/resources/silt density/.