Вы находитесь на странице: 1из 124

1

ASSESSMENT .OF NONLINEAR LEAST-SQUARES


ESTIMATION OF MONOD KINETIC PARAMETERS
FROM BATCH REACTOR EXPERIMENTS

by

MARCO ANTONIO VILLARREAL RODRIGUEZ

Department of Civil Engineering and Applied Mechanics


McGill University
Montreal, Canada

June 1990

A thesis submitted to the Faculty of Graduate Studies


and Research in partial fulfillment of the requirements
for the degree of Master of Engineering

Marco A. Villarreal. 1990


("

( To my parents ....

(
l

"

l'

An academic career compels young man to scientific pro-


i
~1

- duction, and only strong characteres can resist the temp-

tation of superficial analysis.

Albert Einstein.

11
Abstract
1t is known that the use of batch experiments for the estimation of kinetic coefficients

is a better alternative than continuous flow reactors, in terms of the time needed to per-
form t.he experiments.

ln this study a new metl'iodology to e:;t;mate the four Monod-kinetic parameters from

batch reactor data set is presented. This method fits biomass and substrate plots simul-

taneously via nonlinear least-squares analysis. The nonlinear equations are solved via the

secant method.

The methodology performed satisfactorily with two synthetic data sets. It was also

applied to seven batch reactor data sets available in the literature and to ten data sets

from batch experiments using wastewater and sludge from the municipality wastewater

trcatment plant in Granby, Quebec. The constants estimated vary considerably among

thernselves, and sorne were not within the range of values commonly found in the litera-

turc. The methodology proposed yields parameter estimates which produce the minimum

sum of squares residuals.

ft is bclieved that further stuJies are needed if one is to attempt estimating the four

~Iollod-kjnetics constants using batch reactors experiments.

111
Rsum
L'utilisation d'expriences en mode cuve pour l'estimation des constantes de C1n~tiqll('

est reconnue comme tant une meilleure alternative que le racteur coulement continu,

en terme de temps requis pour raliser les expriences.

Dans la prsente tude, ne nouvelle mthodologie pour estimer les <Illettre !>c\l'al1li-t 1 cs

de la cintique de Monod, d'aprs les donnes de racteurs cuve, sera discute. Cctte

mthode lisse simultanment des donnes de biomasse et -le substrat via une analyse

des moindre carrs non-linaires. Les quations non-linaires sont rsolues utilisant la

mthode de la scante.

La mthodologie agit de faon satisfaisante avec deux ensembles de donne:-. .,illluli>(:-,

Elle a aussi t applique sept ensembles de racteurs cuve disponible dan:> la litt{~1 ature

et dix ensembles de donnes provenant d'expriences en mode cuve utilisant des t'(tlIX

uses et des boues de la station d'puration des eaux uses Granby, Qubec Lps (O!,-

stantes estimes varient considrablement entre elles-mmes, et certaines Ile se tl'OllVCllt

pas dans l'tendu des valeurs normalement trouves dans la littrature. Le~ e<;tilllatioll">

de paramtres qui rsultent de la mthodologie propose sont celles qui produisent 1(

minimum des sommes des carrs des rsidus.

Nous croyons que des tudes supplmentaires sont ncessaires si l'on veut tentcr

d'estimer les quatre constantes de la cintique de Monod d'aprs les exprienccs danf>

des racteurs en mode cuve.

IV
Acknowledgements
The author wishes to express his gratitude to Professor R. Leduc, research supervisor,

for his invaluable assistance throughout the research and preparation of this report.

He is also grateful to Professor V.T.V. Nguyen, for his important assistance and con-

structive critlcism.

Special acknowledgemen't is given to Salvador Flores Guerrero (M. Eng.) for his en-

couragement and invaluable advice which made the complet ion of this study possible.

He is particulary indebted to E. Waldron and 1. Buchanan for their important assis-

tance in the environmentallaboratory.

Thanks also go to the authorities of the wastewater treatment plant in Granby, Que-

bec, Andr Charbonneau and Marc Brodeur. for their complete cooperation throughout

the study.

The author was partially supported by the Instituto Tecno16gico y de Estudios Supe-

1 iores de Monterrey, Monterrey N. L. Mxico.

\'
J

Contents

Abstract III

Rsum IV

Acknowledgements V

Contents VI

List of Tables IX

List of Figures x

List of Symbols and Acronyms XIII

1 Introduction 1

2 Literature Review 4

2.1 Wastewater Treatment

2.2 Biomass Growth and Substrate Utilization Cl

2.2.1 Biomass Growth Rate 7

2.2.2 Yield Coefficient . 8

-,
2.2.3 Endogenous Decay 9

2.3 Estimation of Monod Kinetic Parameters . 10

vi
2.3.1 Continuous-Flow Experiments . 10

l 2.3.2 Batch Reactor Expcriments 12

3 Methods and Materials 22

3.1 Model development . . 22

3.1.1 Mathematical analysis 22

3.1.2 Nonlinear least-squares analysis 28

3.1.3 Aigorithm for estimation of the parameters . 30

3.1.4 Estimation of L 30

3.2 Initial Estimates ... 33

3.3 Laboratory Experiments 36

3.3.1 Sampling 36

3.3.2 Ratch reactor 36


.., 3.3.3 Experiments . 37
."
~

4 Results and Discussions 38

4.1 Synthetic Data ...... . .. .. .. . .. 38

1.2 Braha and Hafner (1987) Data Sets 40

4.3 Data from the Granby Plant .. .51

4.3.1 Experiments .. ........ .52

5 Summary & Conclusions 69

A Sensitivity Analysis 72
A.l Intl'Oduction . . . 72

A.2 l'.lodci and Application 73

A.3 Commcnts . . . . . . . 74

l B Main Computer Program 76

Vil
C Non-biodegradable substrate 80

1 ay vs y
D SSRS and ill 83

E SSRX and ffi


aA vs A 85

1
l, F Analysis for the roots 88
~

Vlll
List of Tables

2.1 Typical values oi kinetic coefficients at 20C 10

2.2 Models used by Simkins and Alexander (1984) 18

3.1 Comparison cf methodologies . . . . . . . . . . . . . . . . . . . . . . . .. 32

4 1 Synthetic data . . . . . . . . . 39

4.2 Original and estimated parameters for the synthetic data analysis 39

4.3 Braha and Hafner (1987) Data Sets . . . . . . . . . . . . . . . 42


4.4 Estimated parameters with Braha and Hafner (1987) data sets 43
4.5 Parameters estimated by Braha and Hafner (1987) . . . . . 51

... 6 Characteristics of the Granby wastewater treatment plant. 52

4.7 Characteristics of sludge and substrate for each experirnent . 54

4.8 Estimated parameters for the Granby Plant data sets ... 55

4.9 Ratio of soluble ta total substrate concentration (COD) at Granby. . 55

., 10 Substrate and biomass data for runs 1 to 5 . . . 56

1.11 SSR 's obtained with the Granby Plant data sets 67

A.1 Parameters studied .. 73


A.~ Thirteen smallest values of the SSRs for the first analysis 74

A.3 SSRSs values for the second analysis . . . . . . . .. . 75

IX
'1

List of Figures

2.1 Fl,~W schematic of the suspended-growth activated sludge 5


2.2 Baeterial growth ............ . . . . . 6
2.3 Diomass growth rate vs Substrate concentration 8

2.4 Complete-mixing continuous flow reactor without recycle 11


2.5 Batch Reactor . . . . . . . . . . . . . . .......... 13

2.6 Biomass growth and substrate degradation vs time in a batch reactor 14


''i''l'

.",
3.1 Substrate as the independent variable . 24
3.2 Time as the independent variable 21
3.3 Substrate vs time for different /(IJ 25
3.4 Aigorithm . . . . . . . . 31

3.5 Wand SSRS versus Y . 34


3.6 Wand SSRX versus A 3.5

ay versus y and !l.a


4.1 Synthetic 1: ffi 8A versus A . 40
4.2 8Y versus y and lli
Synthet ie 2: ffi 8A versus A . 41
4.3 SSR ratio, biomass and substrate graphs for LBI 44
4.4 SSR ratio, biomass and substrate graphs for LB2 4.)

4.5 SSR ratio, biomass and substrate graphs for LB3 46


.Jffi;

".1- 4.6 SSR ratio, biomass and substrate graphs for LB4 47

x
4.7 SSR ratio, biomass and substrate graphs for LB5 48

4.8 SSR ratio, biomass and substrate graphs for LB6 49

4.9 SSR ratio, biomass and substrate graphs for LB7 . .... 50

4.10 SSR ratio, biomass and substrate graphs for run 1 (COD) 57

4.11 SSR ratio, biomass and substrate graphs for run 1 (TOC) . 58

4.12 SSR ratio, biomass. and substrate graphs for run 2 (COD) 59

4.13 SSR ratio, biomass and substrate graphs for run 2 (TOC) . 60

4.14 SSR ratio, biomass and substrate graphs for run 3 (COD) 61

4.15 SSR ratio, biomass and substrate graphs for run 3 (TOC) . 62

4.16 SSR l'atio, biomass a:d ",ubstrate graphs for run 4 (COD) 63

4.17 SSR ratio, biomass a'1d substrate graphs for run 4 (TOC) . 64

4.18 SSR ratio, biomass and substrate graphs for run 5 (COD) 65

4"
4.19 SSR ratio, biomass and substrate graphs for run 5 (TOC) . 66
'i
F.l -W- vs Y, SSRS vs Y, ~7 vs A and SSRX vs A for LBI . 89
F.2 -W- vs Y, SSRS vs Y, ~7 vs A and SSRX vs A for LB2 . 90

F.3 -W- vs Y, SSRS vs Y, ~7 vs A and SSRX vs A for LB3 . 91

F.4 -W- vs Y, SSRS vs Y, a:: vs A and SSRX vs A for LB4 . 92

F.5 -W- vs Y, SSRS vs Y, ~7 vs A and SSRX vs A for LB5 . 93

F.6 ~ vs l', SSRS vs Y, a:: vs A and SSRX vs A for LB6. 94

F.7 * vs Y, SSRS vs Y, a:: v!: A and SSRX vs A for LB7 . 95


F.8 * vs Y, SSRS vs Y, ~7 vs A and SSRX vs A for Run 1 (COD) 96

F.9 * vs Y, SSRS vs Y, ~7 vs A and SSRX vs A for Run 1 (TOC) 97

F 10 ~ vs Y, SSRS vs Y, ~7 vs A and SSRX vs A for Run 2 (COD) 98

F.ll * vs Y, SSRS vs Y, ~ vs A and SSRX vs A for Run 2 (TOC) 99


f F.l2 W- vs Y. SSRS vs Y, ~ vs A and SSRX vs A for Run 3 (COD) 100

Xl
.j

F.13 ~~ vs Y, SSRS vs Y, ~~ vs A and SSRX vs A for Run 3 (TOC) 101

F.14 ~~ vs Y, SSRS vs Y, fi vs A and SSRX vs A for Run 4 (COD) 102

F.15 ~~ vs Y, SSRS vs Y, fi vs A and SSRX vs A for Run 4 (TOC) 103

F.16 ~~ vs Y, SSRS vs Y, fi vs A and SSRX vs A for Run 5 (COD) 104


F.17 ~~ vs Y, SSRS vs Y, fi vs A and SSRX vs A for Run 5 (TOC) 105

xii
List of Symbols and Acronyms

A constant
B constant
COD chemical oxygen demand
F/M food-biomass ratio
k maximum substrate utilization rate per unit of biomass
kd endogenous decay coefficient
c d estimate of endogenous decay coefficient
kt constant
k2 constant
k3 constant
k4 constant
1<1 first-order rate constant
L constant
L fanal upper limit of L
Lo lower limit of L
Lp value of L when the plateau is reached
mznSSR minimum sum of squared residuals
.,. MLVSS mixed liquors volatile suspendid solids
,.
:;t
q inverse yield
Q influent flow rate
QR recycle line flow rate
Qw sludge wasted flow rate
rd endogenous decay rate
rg biomass growth rate
.
rll
rg
substrate utilization rate
biomass growth net rate
R by-products from substrate consumption
S growth-limiting substrate concentration
Se biodegradable substrate concentration at the end of the experiment;
effluent biodegradable substrate concentration
S, biodegradable substrate at time tj
So biodegradable substrate concentration at the begining of the experiment;
influent biodegradable substrate concentration
St pl'edicted substrate concentration at time t

{
xiii
l

SSR sum of squared residuals


SSRS sum of squared residuals of biodegradable substrate concentration
SSRX sum of squared residuals of the logarithm of biomass concentration
SVl sludge volume index
S average biodegradable substrate in the time interval 6t
t time
ti time at the i th observation
TOC total organic carbon
V volume
Vm maximum substrate consumption
X biomass concentration
Xa average biomass concentration
Xo biomass concentration at the begining of the experimentj
influent biomass concentration
XR biomass concentration recycle tine
X average biomass in the time interval D.t
."..,..
X* substrate concentration required to produce a biomass concentration X
J...
X* 0 substrate concentration required to produce a biomass concentration X o
y total yield coefficient
'Y constant
D.[(s correction term for half-velocity constant
D.S biodegradable substrate depleted in the experiment
D.t time interval
D.X biomass growth in the experiment
D.Y correction term for the total yield coefficient
D.p.m correction term for the maximum specifie growth rate
() hydraulic retention time
J.l specifie growth rate
J.lm maximum specifie growth rate
e constant
4> constant
w constant

XIV
Chapter 1

Introduction

One of the major princip les limiting the formation of human communities has been the

preoccupation with securing and maintaining a suitahle supply of clean water. Unfortu-

nately, water has played an important role as a carrier of wastes throughout the history

of civilizations.

It is known that over 100 different types of viruses and pathogenic bacteria may

he present in untreated domestic wastewater. (Tchobanoglous and Schroeder, 1985 [29]).

The first documented relationship between human wastes, drinking water, and disease

was reported in the mid-nineteenth century. Nevertheless waterborne disease was not

properly accepted un tH more advanced studies in the germ theory of disease were done

by Pasteur and other scientists at the end of the nineteenth century and early twr>ntieth

cent ury (Montgomery, 1985 [14]). It is now known since many decades that microorgan-

isms can be responsible for specifie public health prohlems mcluding bacterial illnesses

such as cholera and gastroenteritis, viral diseases such as hepatitis, amoebic dysentery or

diarrhca which are generated by protozoa, and parasitic helminth (worm) infections such
{ as tapeworms or roundworms (Montgomery, 1985 [14]).

1
1 The first goal for the water treatment professionals in the nineteenth cent ury was

to eradicate waterborne diseases. This was accompli shed successfully in the developed

countries by the virtual elimination of the mos t deadly waterborne diseases such as ty-

phoid, cholera and amebiases. However, since the beginning of the twentieth ccntury,

public health concerns have switched from acute disease to the chronic health effects of

trace quantities of organic, inorganic and microbiological pollutants.

On the other hand, wastewater carries wastes of different types from different sources

including residential, commercial, industrial wastes, ground water, surface water, stormwa-

ter, etc. These \Vastes need to be removed for pollution control before the watel" is dis-

charged into the receptor body (river, lake, etc.). The discipline that deals \Vith \Vastew-

ater treatment is called wastewater engineering.

Wastewater engineering is a branch of environmntal engineering which applies the

basic principles of science and engineering to the problems of water pollution control.

The ultimate goal of this engineering branch is to develop wastewater management tcch-

niques in order to help protecting the environment. Physical, chemical and biological

pro cesses are generally involved in a municipal wastewater treatment plant. Disinfection

and coagulation are examples of chemical processes, while sedimentation, filtration and

screening are examples of physical processes.

Biological treatment processes are a significant and indispensable segment of domestic

wastewater treatment plants. In particular the activated sludge process is extcnsively

used, mainly due to its efficiency in degrading organic matter and suspended solids. The

activated sludge process was deve!oped in England in the carly 1900s (Joint Committee of

2
----------------~----------------------------------- -

the WPCF and the ASCE, 1977 [23]) and was brought into North America in the 1920s.

1 By the 1940s the process has been used throughout the world.

AlI the biological pro cesses present in a wastewater treatment plant take place in

a defined volume. Sueh a volume is eommonly named a reactor. Important changes in



the composition and eonce~tration of organic eompounds occur while the wastewater is

retained in the reactor. These changes are caused by hydraulic transport of materials into

and out of the reactor as well as by reaetions that oeeur within the reaetor. To completely

define a reactor system and design similar ones it is necessary to know the rate at which

the change~ of composition and concentration of biodegradable substance occur and the

extent of the changes.

To design any biological pro cess for a wastewater treatment faci li ty, it is required

to know the rates at whieh various components (such as organic compounds) are de-

graded from the wastewater and the rate at which biomass is produeed in the reactor.

Sueh rates of change are important because they will determine the size of the reactor

required for a specifie degree of water quality.

The most weil known mathematical mode! which describes the rates for the degra-

dation of organic material and the biomass produced in a bateh reactor is the Monod

kinetics model (Monod, 1949 [20]). This model includes four parameters ~hat need to

be estimated to properly design biological units for wastewater treatment. In this study

a new methodology is developed to estimate the parameters involved in Monod kinetics

using various sets of data from bateh reactors.

3
.J

Chapter ~

Literature Review

2.1 Wastewater Treatment

Wastewater is classified into industrial and municipal. The characteristics of inctust.l'ial

wastewater vary with respect to the type of industry, therefore its treatment will be (liffer-

ent for each case. However the characteristics of municipal wastewater are bettel' kllowlI

because they are affected mainly by domestic uses, relathely small hydraulic and organic

loads from industries (with respect to domestic loads), stormwater, infiltrations, among

others. Domestic wastewater contains mainly biodegradable organic matter, suspended

solids, pathogenous microorganisms, nutrients, and a wide variety of compounds in lllllCh

smaller concentrations_

Biological treatment pro cesses are generally involved in domestic wastewatcr treat-

ment plants. In biological processes, microorganisms (mainly bacteria), present in the

wastewater, convert easily biodegradable organic pollutants into degraded products Thf!

biological process can be classified, according to the type of baderial growth, iIlto at-

tached or suspended-growth.

In the attached-growth process a medium is utilized to support the microorganism~.

4
------------------------

This growth process is carried out usiog trickliog filters, rotatiog biological contactors,

activated biofilters, among others (Benefield & Randa1l1980 [3]).

One of the most common suspended-growth processes in wastewater treatment plants

is the activated sludge .. The system comprises three components, which are the aerated re-

actor, the clarifier and a sludge recycle line (see Fig. 2.1). In the reactor, the biodegradable

Clarifier
Reactor
(

Recycle line

Figure 2.1: Flow schematic of the suspeoded-growth adivated sludge

organic matter is degraded by the biomass under appropiate conditions (pH, temperature,

oxygen, etc.). The biomass grows and forms ftock that oormally settle in the clarifier.

The sludge (i.e. biologicai solids, water and other dissolved constituents) is withdrawn

fl'om the bot tom of the clarifier. A portion of that sludge is returned to the aeration tank

and the rernailling is wastcd. As an alternative, sludge can also be withdrawn from the

aeration basin.
i
5 1

j
'J

2.2 Biomass Growth and Substrate Utilization

Effective environmental control in biological wastewater treatment is based on an lIllder-

standing of the basic principles govern',ng the growth of microorganisms. The gt'Ilcl"al

growth pattern of bacteria in a batch culture is shown in Fig. 2.2. In aCl'obic batch cul-

ture experiments, bacteria are provided with enough soluble food (substrate) and suitabk
.
environmental conditions (pH, temperature, redox, and others).

The growth of microorganisms under aerobic conditions in batch reactors can be

Cl)
II) en Q)
CI) Cl !Il >-
.... CI)
<Il C s::. <Il co
s::. c a. c Cl>
s::.
a. 0
<il
C.
>-
-cin Cl)
CI)
<Il
II)
'0
-
s::.
~ <Il
<Il
s::.
C.
CI)
::l
!Il
s::.
Q) 0
.... <il 0
c
OJ
Q)
"C
a. c Q)
0 Cl
c a. 0
<Il
.... (1) -0 x V
'\
"C
C

/
CI) Q)
!Il
II)
<Il ....

~
s::.
.... -
C,) Cl)
<Il c. (l)
8 Cl (.)

Cl ;:: (.)
ttI
~

/
Tlme t

Figure 2.2: Bacterial growth

divided into six phases (Benefield & Randall 1980 [3]) (see Fig. 2.2).

1. The lag phase is characterized by the adaptation of microorganisms to thcil'

newenvironment. There occurs log generation time and null growth rate, and

cell size and rate of metabolic activity are at maximum.

2. During the acceleration phase there is a decrease in generation time and an

increase in growth rate.

3. The exponential phase is characterized by maximal and constant specific gl'Owth

rate, maximum rate of substrate conversion and constant gcneration time.

6
4. The declining growth phase exhibits an increase in generation time and a de-

crease in specifie growth rate due to the graduaI decrease in substrate concen-
l tration and the incased accumulation of toxic metabolites.

5. In the stationary phase the nutrients are exhausted, leaving behind a high

concentration of toxie metabolites, and maximum physical crowding .



6. In the endogenous phase there is endogenous metabolism, high death rate, and

cell lysis takes place.

2.2.1 Biomass Growth Rate

The rate of growth of biomass during exponential growth (phase 3), can be described by

the following equation:

(2.1)

where

Tg = biomass growth rate, mass vo/ume- 1 time- 1

Jl = specifie growth rate, time- 1

x = concentration of microorganisms, mass volume- 1

In the case when the biodegradable soluble substrate is growth-limiting, the specifie

growth rate can be described using Monod kinetics (Monod, 1949 [20]):

(2.2)

where

ILm = maximum specifie growth rate, time- 1

S = concentration of growth-limiting substrate, mass volume- 1


/(5 = half-velocity constant, mass vo/ume- 1

7
The relationship between the specifie biomass growth rate and the limiting substrate

concentration is presented in Fig. 2.3.


1 lntroducing Eq. ( 2.2) in ( 2.1) gives:

IlmXS
rg = [(6 +S (2.3)

fJ-m ........... _.................... _- ............. __ ................ _...... -- -. _ ............... _-_ ... -......... -_ ........ .

fJ-m
2

Substrate

Figure 2.3: Biomass growth rate vs Substrate concentration

2.2.2 Yield Coefficient

In aerohic culture systems, a portion of the substrate is converted to new cells and the

other portion is oxidized to end products. The quantity of new cells produccd has bccn

observed to be a function of the substrate removed. The following relationship has bccn

developed between the rate of substrate utilization and the rate of biomass growth:

(2.4 )

where

y = total yield coefficient, mass mass- 1


r6 = substrate utilization rate, mass vo/ume- 1 time- 1
8
The total yield coefficient Y refers to any fini te period of exponential growth, and is

defined as the ratio of t.he mass of cells formed to the mass of substrate consumed.
l Introducing Eq. ( 2.3) in ( 2.4), we have:

J.lmXS
(2.5)
r" = Y(K. + S)
Setting k = ~ gives:

kXS
(2.6)
r" = [(" + S

where k represents the maximum substrate utilization ra.te per unit of biomass, (time- 1 ).

2.2.3 Endogenous Decay

In bacterial systems used for wastewater treatment, the distribution of cell ages is su ch

that not aIl the cells in the system are in the same phase. The microorganisms which are

in the endogenous phase do not normally generate new cells, therefore inducing a decrease

in the growth rate called endogenous decay.

Endogenous decay not only takes into account the cells age but also the energy of

maintenance and predation (Herbert, 1958 [13]). The rate of endogenous decay (rd) has

been found to be proportional to the concentration of microorganisms in the system and,

hence, it follows the relationship:

(2.7)

where kd represents the endogenous decay coefficient, (time- 1 )

The net rate of bacterial growth (r~) is written:

(2.8)

Introducing Eqs. ( 2.3) and ( 2.7) in ( 2.8) produces:

(2.9)

9
Table 2.1: Typical values of kinetic coefficients at 20C
Coefficient Units Range
Benefield & Randall (1980) Metcalf & Eddy (199)
day-l 6-S 2 - 10
mg/l COD 25 - 100 15 - 70
mg MLVSS
mg ('OD 0.35 -0.45 0.25 - OA
day-l 0.05 - 0.1 0.04 - 0.075

where r~ is the net rate of b~cterial glOwth, (mass volume- 1 ttme- 1 ). Thc range of values

commonly found for Y, ](a, k, and kd for municipal wastewater are presented in Table 2.1.

2.3 Estimation of Monod Kinetic Parameters

The general characteristics of domestic or municipal wastewaters are weil known. Ncvcl'-

theless, for each case an estimation of the parameters (i.e. the biokinetic coefficlellt) i~

.'
strongly recommended for the design and operation of a wastewater treatment plant is

planned (Vasicek 1982 [34]).

Values for the parameters Y, p'm, Ka and kd must be found for a particulal' case and
used in the design for most effective results. Two methods can be consideree! to detcrrnnc

these coefficients: continuous-flow systems or batch reactors.

2.3.1 Cont inuous- Flow Experiments

To estimate the parameters, the usual procedure with continuous-flow reactrs is t op er-

ate the units over a wide range of retention times (i.e. flowrates), and influent cO!1ceIlLll

tions (5 different retention times ranging from 1 to 10 days are recommcnded by BCllcfidd

and Randall1980 [3] and Met.calf & Eddy 1979 [19]). When steady-statc l~ Icac!l('u fol' t

specifie retention time, the v.ilues of discharge (Q), substrate concentration in thc illnucllt
.'
(Sa), substrate concentration in the effluent (Se), and microorganism!> conccnt.ration in

10
----------~----------------------_.----

l Se , Q , X
- ---

v, x, Se

Figure 2.4: Complete-mixing continuous flow reactor without recycle

the tank (X) are measured.

Applying subc;trate balance to the system and assuming steady-state, the following

equation is obtained:

(2.10)

realTanging ylelds:

(2.11)

\Vriting Sr = So - Se, and () = ~ where () is the retention time (time), this last equation

becomes:
....'
rs = -"r (2.12)
o
Equating Eqs. ( 2.12) and ( 2.6), and linearizing, the following relationship is obtained:

OX = (1(,) ~ + ! (2.13)
Sr k Se k
Dy plotting ~~ vs i. a straight \ine with slope If' and intercept i is obtained.
FUl'thermore, a stcady-state mass balance on the system can be written assuming that

active biomass in the influent is negligible, as:

11
(2.14)

1 Introducing ( 2.9) and ( 2.12) in this last equation, and rearranging, gives the following

linear relationship:

1 Sr
- = y - - kd (2.15 )
() ()X
""
Plotting ~ vs fff gives etimates of Y and kd'

2.3.2 Batch Reactor Experiments

The use of continuous-flow experiments for parameter estimation, involves a considerable

amount of time. It is known that to achieve steady-state in the system, apPl'oximn.tely

two to three weeks are needed after the initiation of the test. Consequently, if diffcl'cnt

retention times are required [Benefield and Randal (1980) [3], and Metcalf & Eddy (199)

[19]] for processing the data, either two to three rnonths will be necessary to complete

the study, or sever al reactors should be run in parallel which implies significant amount

of qualified personnel and equipment.

On the other hand, batch reactors are closed systems in which a specifie concentration

of adapted biomass is allowed to grow in the presence of substrate under a suit,ablc Cllvi-

ronment (pH, oxygen, temperature, etc ... ) as shown in Fig. 2.5. The reaction is genpI'ally

allowed to proceed until the biodegradable substrate is almost completcly dcpldcd by

the biomass. During the process, substrate and biomass concentrations arc mcasUI'cd at

different times. Generally, the experirnent will not last for more than 24 hr when adaptcd

biomass is available. If no adapted biomass is available, there is a nccd t obtaill ~()me

using a batch, or continuous flow reactor.

Fig. 2.6 shows biomass growth, and substrate disappearance as a functioll of time

where So, X Ol Se and Xe indicate the substrate and biomass conccntration at time zero,

12
VI XI S

Figure 2.5: Batch Reactor

and substrate and biomass ccn~entration at the end of the process, respectively.

An essential element in any comparison between batch and continuous-flow experi-

ments, is that batch reactors represent the best alternative for biokinetics estilnation in
( ter ms of the time needed to obtain the parameter estimates. Few methodologies has been

developed to estimate the Monod biokinetics parameters using batch reactor experiments.

Most of them apply numerical methods while others employ either linearization, graphi-

cal, chemical analysis or respirometric measurements. These methodologies are reviewed

hereafter.

(a) Reynolds and Yang (1966) Method [25]

This is an introductory methodology in which the authors used five batch reactors

simultaneously containing different substrate and biomass concentrations. The authors

measured substrate and biomass concentration at time zero and at time 24 hr. in each

flask l'eactor. In this study the equations were rearranged by introducing Eqs. ( 2.4) and

( 2.7) in ( 2.8), yielding:


(
(2.16)

13
kd = 0
............. ........ X
. '
.' . .. '

s.
Time t
Figure 2.6: Biomass growth and substrate degradation vs time in a batch reactor

Discretizing Eq. (2.16) gives:

(2.17)

where 6.X is the biomass growth (Le. Xe - X o ) mass volume-l, b.S is the substrate

depleted (i.e. So - Se) mass volume-l, and b.t = duration of the experiment.

Setting b.t = 1day and X = X a where X a is the average biomass during the experi

ment, [i.e. X a = !(Xo + Xe)] and rearranging yields:

(2.18)

Plotting ~: vs ~: provides estimates of Y and kd'

(b) Robinson and Tiedje (1983) [27]

In this study a nonlinear regression analysis was proposed to estimate Monod kinetics.

Robinson and Tiedje used the Monod integrated form for substrate removal [Eq.( 2 ..5)],

in which biomass was eliminated in Eq. ( ~.5) by using the following relationship:
....
(2.19)

14
Introducing Eq. ( 2.19) in ( 2.5) yields:

( dS Jl.mS [Xo + Y(So - S)]


(2.20)
dt Y(I(3 + S)

Performing the integration they established:

(2.21 )

Robinson and Tiedje discerned that high correlation existed among the variables, and

argued that a nonlinear least-squares technique would not be the appropriate routine to

use, although from the statistic point of view it is considered the best alternative [Bach

(1982) [2]].

The authors proposed instead a nonlinear regression analysis using the foIlowing model:

dS dS dS
St = ~Pm dPm + tlI<3 dK + tlY dY
3
(2.22)

where the derivatives were solved by using implicit differentiation in Eq. ( 2.21).
{
The analysis was carried out using a computer program (MONODCRV) that estimates

the theoretical predicted substrate (St) at time t with Eq. ( 2.22) (given initial estimates

of the parameters) and calculate the residual errors (S - St) for aIl the points, where S

is the observed substrate at time t. tlpm, tlI<s and ~Y were solved via multiple linear

regression and added (either positive or negative) to the initial values. The process con-

tinued until the correction terms were less than 0.01. The initial guesses of the parameters

were evaluated with the linearized forms of Eqs. (2.3) and ( 2.6) which are discussed

hereafter. The model was tested with success using different sets of simulated data and

one set of H 2 -1imited batch growth of Desulfovibrio spp.

The authors defined their method as a nonlinear regression analysis with Gaussian

mcthod. Nevertheless, it is considered that this methodology is rather a combination of

the Newton Raphson and the steepest descent methods.

15
1
.J

(c) Robinson and Characklis (1984) Method [26]


In this study the authors estimated the parameters from the integrated form of the

Michaelis-Menten equation. This equation was used to describe substrate depletion ac-

cording ta the foIIowing relationship:

(2.23)

where

Vm = maximum substrate of consumption, mass volume- 1 time- 1


R = by-products from the substrate consumption, mass volume- 1 time- 1
The model did not consider biomass growth.

The authors estimated the parameters by using Eq. (2.23) in four different fonus:

one nonlinear and three different linearized forms. In the nonlinear approach the authors

used the same nonlinear regression analysis applied by Robinson and Tiedje (1983) [27J,

while they used linear least-squares analysis with the linearized forms.

Robinson and Characklis (1984) [26J found that applying nonlinear regression analysis

ta the integrated form and considering So ta be not free of error, yielded bcttcr results

than any of the other three Iinearized forms by using the !inear least-squarcs technique.

(d) Simkins and Alexander (1984) [28]

The authors introduced a new mass-balance equation as follows:

So + qXo = S + qX (2.24)

where q = is the ceU quota or inverse yield (~) and X o is the initial biomass concentration.

Setting qX equal ta X*, which corresponds to the amount of substrate required ta produce

a biomass concentration equal to X. In the same manner X; is defincd as bcing equal ta

....... qXo. Rearranged, Eq. ( 2.24) can be written as:

(2.25 )

16
Solving for X* from Eq. (2.25) and its derivative with respect to time, and introducing

1 both terms into the differential form of Eq. ( 2.3) yields:

_ dS = /-lmS (S X* - S) (2.26)
dt Ka +S 0 + 0

The authors employed Eq. (2.26), rearranged for a particular set of assumptions. The

assumptions and the respective models are presented in Table 2.2. Data were collected

during an experiment on the mineralization of benzoate byan induction of Pseudomonas

spp. The non!inear regression analysis was carried out using MARQ-FIT, which performs

a minimization of least-squares of data by the method of Marquardt (1963) [18].

In this study, all the models produce positive results, excluding the model where no

assumptions were stated\ i.e. the Monod model with growth. For this set of data the

Jogistic mode! yielded a better fit than the Monod model with growth, eventhough the

value of So was close to that of /(3' The authors offered no possible explanation for that.

If we examine the high correlation among the parameters, as pointed out by Robinson and

Tiedje (1983) [27], and Robinson and Characklis (1984) [26], the parameter correlation

rises as a possible explanation.

(e) Braha and Hafner (1987) Method [7]

The authors presented a new methodology to evaluate Y, K31 /-lm, and kd , while

the other studies (presented above) have only estimated two or three out of the four

parameters. In this study, the authfs utilized the methodology proposed by Rey 10lds

and Yang (1966) [25]. They plotted the data from a batch reactor using Eq. ( 2.18), to

obtain the values of Y and kd through !inear least-squares. The two other parameters (1<3

and /lm) were found through nonlinear least-squares analysis of Eq. ( 2.21) (by using a

computer program the BP7IB application of the curve-fit module).

The use of Reynolds and Yang (1966) [25] methodology in a single batch reactor
l makes this technique a questionable method, mainly due to the units of kd' In their

17
.1

,
1
Table 2.2: Models used by Simkins and Alexander (1984)
ModeZ and characteristics Equations and inequalities
1. Zero Order
Differentiai form -dS - k
dt - 1
Integrated form S = Sa - kIt
Derived parame\er kt = JlmX;
Necessary conditions X; ~ So and So ~ J(,

II. Monod, no growth


Differentiai form -dS _ .J!J.L
dt - K.+S
Itegrated form [(,ln 10 + S - So = -kIt
Derived !>arameter kI = JlmX~
Necessary condition X; ~ So

III. First order J


Differentiai form -dS -
dt -
k3 S 1/,
Itegrated form S = Soe- k3t ~

Derived parameter k - ~"'x~ il


3 - K.
..- Necessary conditIOns X; ~ So and So J(, 1

il'
IV. Logistic
DifferentiaI form -fts = k4 S(So + X; - S)
Integrated forrn S - so+x;
- I+4e"4(50+X~)1
50
Derived parameter k4 -~
- K.
Necessary condition So !(,

V. Monod with growth


-dS _ IJmS(So+X~-S)
Differen tial form dt - K.+S
Integrated form [(,In la = (Sa + X; + K,) ln f~ - (So + X;)llm t
Derived parameter None
Necessary condition None

VI. Loga7'ithmic
DifferentiaI form -fts = J1.m(So + X; - S)
Integrated form S = So + X;{l- elJmt )
Derived pararneter None
Necessary condition So ~ !(,

l
18
original study [Eq. ( 2.18)] the units of kd are day-I because the measurements were done

24 hours after initiation of the test, while Braha and Hafner measured the substrate and
1 bioma&-; concentration throughout the trial (and contemplated aIl the points). If time is

considered in Eq. ( 2.17) i.e.

I:J.X =y t:J.S _ k
X a t:J.t X a t:J.t d

and the methodology is applied, the estimated values of Y and kd are negative.

(f) Van Niekerk et al, (1987) [33]

These authors model the soluble COD uptake in batch reactor using the foIlowing equation

(2.27)

where [(1 is the first-order rate constant, volume mass- 1 time-t, estimated via linear

regression analysis.

The authors pointed out that first-order substrate uptake can be regarded as the sum-

mation of a number of zero-order removals [Tischler and Eckenfelder (1969) [31]].

(g) Templeton and Grady (1988) [30]

In another study, Templeton and Grady (1988) used Eq. ( 2.21) to estimate Ks, JLm

and Y. The parameters were obtained with Marquardt's B-Solve algorithm [Marquardt

(1963) [18], Kuester and Mize (1973) [15]]. The authors used no other model or technique

to do a comparative analysis.

(h) Discretization and linearization techn~ques

The discretization and Iinearization of equations for parameters estimations is a corn-

mon technique used by chemical and environ mental engineers. Renee, by discretizing and

!inea! izing Eqs. ( 2.3) and ( 2.6) the following relationships are obtained:

19
~tX J(.Y 1 Y
---=--=+- (2.28)
~S J-lm S /lm

!ltX J(. 1 1
-=--=+- (2.29)
!lX /lma.x S J.lma.x

where 6..X, IlS, X and S are the biomass growth, the substrate depleted, and biomass
and substrate averages respectively in the time interval 6..t. If 6..t is rclativcly small.

plotting -~: and ~t: ver~us ~,estimates of [(., J.lm and Y are obtained (Robinson and

Tiedje (1983) [27]). Therefore, with a set of data of substrate and biomass versus lime

from a batch reactor, it would be possible to find the constants f(., JLm and lI" (but not kd ).

However several authors have found vulnerable points concerning the lincarization of

equations. Robinson and Tiedje (1983) [27], and Robinson and Characklis (1984) [26]

found that estimating parameters by nonlinear regression analysis generates constants

which fit the data points better than by discretization and linearization techniques. 8raha

and Hafner (1987) [7] pointed out that by-products were generated by the microorganisms

which would affect bacteria growth in a closed system, and that these cffects WCI C Ilot

considered in Eqs. ( 2.28) and ( 2.29).

In summary among the methodologies commented above, only the method by Araha

and Hafner (1987) [7] yields estimates for al! the constants. Yet, this method is qlles-

tionable technique (as explained above and proved in Appendix A). Consequently, if ail

estimation of the four Monod parameters is required, a new methodology is desirablc On

the other hand, it was noticed in the study by Sirnkins and Alexander (1984) [28], that

\Vhen no assumption was made in Eq. (2.26), the pararnetcrs estimated did Hot yicld the

minimum sum of squares residuals (SSR), [the logzstic modcl (k<1 = y.) yielded smallcr

SSR]. It is unknown as to why the computer packages converge to wrong estimates iu

20
the two studies commented above, but from this analysis it is strongly recommended to

perform a sensitivity analysis similar to the one carried on in Appendix A if a computer

package is applied.

Finally it is not th scope of this thesis to coyer either graphical, chemical nor respiro-

metric techniques for parameters estimation sinee this study focusses on numerical ap-

proaches only. However it is important to mention herein that only the methodology

proposed by Blok (1974 [5] and 1976 [4])(respirometric measurements) estimates the four

parameters. On the other hand, Camara and Randall (1984) [9] (chemical) and Gates

and Marlar (1968) [10] (graphical), estimate one and two parameters, respectively.

21
1

Chapter 3

Methods and Materials

In Chapter 2 it is pointed out, that the approaches for estimating the biokinetics yielded

only two or three out of the four Monod parameters. The only rnethodology which yiclds

the four constants was proposed by Braha and Rafner (1987) [7], but it was showll in

Chapter 2 that it is a questionable methodology. Therefore, a new model should he

developed if one attempts to estimate the four Monod parameters from batch reactor

experiments.

In this study a new methodology is proposed and assessed to estimate the f01l1' pa-

rameters of the Monod kinetics expression using various sets of data from batch reactors.

The methodology presented herein fits simultaneously biomass and substrate data sctl>

through nonlinear least-squares analysis.

3.1 Model development

3.1.1 Mathematical analysis

(a) Substrate Utilization


The rate of substrate utilization is defined by Eq. ( 2.20) as follows:

22
dS J1.mS [Xo + Y(So - S)]
-=- (3.1 )
dt Y(I<.. + S)
l Rearranging and integrating Eq. ( 3.1) yields:

J1.
1
m 0
t
dT - -
-
1SSo
K 6 Yd
[Xo + Y(So - )]
-
i
50
5

Xo
Yd
+ Y(So - )
(3.2)

Performing the integr~tion of Eq. ( 3.2) gives:

(3.3)

Substituting the upper and lower limits for substrate and rearranging, yields the following

equation:

(3.4)

Eq. ( 34) is the same as the one used by Robinson and Tiedje (1983) [27], Braha and

Hafner (1987) [7], Braun and Berthouex (1970) [8] and Moser et al. (1973) [21] in their

studies [i.e. Eq. ( 2.21), but arranged differently].

The least-squares techniqae considers the independent variable as being error-free, and

minim.ize1> the sum of square residuals (SSR) with respect to the dependent variable. In

Eq. ( 3.4) least-squares assumes that the measured substrate is considered to be error-free

and the time variable is assumed to contain the source of error (see Fig. 3.1). Bence,

to properly apply the least-squares technique, time must be considered as error-free (i.e.

the independent variable) and substrate the source of error (see Fig. 3.2).

Analyzing the plots of ex peri ment al data presented in the studies by Simkins and

Alexander (1984) [28] and Braha and Hafner (1987) [7], the relat.ionship between l\~ and

80 cau be Jtablished from a plot of substrate degradation versus time for Ks > So,

[{~ ~ So and !(! < So. These relationships are shown graphically in Fig. 3.3.
Uloth and Mavinic (1977) [3:!] and Lawrence and McCarty (1970) [17], pointed out

23
s
1 0 model

0 observation

Cf) er ror
....co
<J)

....
CI) -----0
.0
::J
Cf)

0----

Time t

Figure 3.1: Substrate as the independen t variable

so
model

o observation

(J) er ror
CD
.......
co
"'-
CI) o
.0
::J
(J) o
o

Tlme t

Figure 3.2: Time as the independent variable

l
24
(f)
<D
co
~

(f)
.0
:J
(f)

K8 ) S 0

Time t
Figure 3.3: Substrate vs time for different /(~

that as the complexity of the waste increases, the value of /(5 increases. Hence, consid-

ering that the samples used in our experiments faIl into this group (since two industries

discharge their wastes into the sewage of Granby), and looking at the shape of each of the
( plots of our lahoratory experiments (which are similar to the shape for Je > So , Chapter
4), the value of l<S is considered to be larger than So, and the second term of the right

hand side of Eq. ( 3.4) [i.e. _1 ln Y(So1 l+Xo ] hecomes negligible and therefore it can he
/.lm
eliminated, allowing time to be the independent variable.

From the above assumption Eq. ( 3.4) C:]T) he simplified to the following form:

t = y l(, ln So[Y(So - S) + Xo]


(3.5)
JLm(YSo + X o) XoS
or

(3.6)

where L is a constant defined as

(
25
Furthermore, if the sarnple tested has /<, close to So, this model could still be applied. In
this case, an inapropriate fit would occur when time is small (i.e. when filting the first

points). Therefore sorne accuracy would be lost in that portion of the curve, but the l'est

of the points would have an appropriate fit .


(h) Biomass Growth

In Chapter 2 the cell growth was established by Eq. (2.9), which considcl's endogenolls

decay. Herein this equation is used in its differential form, Le.

dX _ l'm SX _ k X (3. )
dt K, +S d

If substrate is elirninated in Eq. ( 3.7) by using the following relationship S' = So + .\\-:- \"
[see Knowles et al (1965) [16], Pirt (1975) [24] and Robinson and Tiedje (1983) [27]}, a

new equation is established:

_dX _ l'm X (YSo + X o - X) _ kdX (3.8)


dt y (I<, + So) + X o - X

Performing the integral for values of X o at t = 0 and X at t = i, the following cquation


is obtained:

(:3.9 )

where w and 1 are constants defined as:

(:l.1O )

(:3.11)

In Eq. ( 3.9) the variable time is expressed as a dependent variable and it cannot he ar-

ranged to be independent. Therefore, to establish time as an indepcndent variable sorne

26
assumptions may be considered in the analysis.

l If /(6 is considered to be much larger than So, then Scan be eliminated from the

denominator in Eq. ( 3.7), thus creating the following relationship:

dX LX
-dt= - (yY So.+X0 -X)-kdX (3.12)

Effecting the integration and rearranging generates the following equation:

e4>t
X - ----'-,.----- (3.13)
- 1 + ..L
X
- e tPt
o

where ~ and <p are constants defined as:

(3.14)

(3.15)

By using Eqs. ( 3.6) and ( 3.13), it is not possible to estimate the four parameters due to

the fact that the respective values of /lm and /(6 cannot be estimated separately, but only

the value of the ratio L = 71;, [this was acheived in the ~tudy of Simkins and Alexander

(1984) [28]].

An alternative to the above approach consists in substituting the substrate expression

Eq. ( 3.6) in Eq. ( 3.7). This yields the following relationship

(3.16)

where A and 0' are constants defined as:

(3.17)

27
1
Performing the integration of Eq. (3.16), and considering X o at t = 0 and X at t = t, givcs

ot
ln (~) = (B _ kd) t _ YS oin (A+xoe ) (3.18)
Xo A A A+Xo

It was ohserved by inspection in Eq. (3.18), that when t = 0, X = X o and when

t -+ 00, X -+ o.

3.1.2 Nonlinear least-squares analysis

Nonlinear least-squares analysis was performed using Eqs. (3.6) and ( 3.18) for substratc

removal and ceU growth respectively. For the purpose of simplicity, X 0 and So were

considered in this study to he error-free (i.e. the readings for X o and So wele considered

as one hundred percent accurate). This assumption probahly affects in somc llIallllCr the

values estimated, but it is believed that this simplification is appropriate at tlli" ~tagc in

the development of this new methodology.

Rearranging expression ( 3.6) for least-squares analysis yields:

(3.19)

where n is the number of substrate-time observations.

Performing the derivative of Eq. ( 3.19) with respect to Y gives:

(YSo + Xo)2(So _7P~L(So+~)t.)] + SoS,(l'So + Xo(So+~Jt.)2


{ (Y So + XoeL(So+~Jt')3
_ (l'So + Xo)[S; + S,(So - L;~t. eL(So+~Jt.)] }
..:..-.----.:..:..-=---~----tx------'-'- (:3.20)
(YSo + x o(So+-v Jt ')2
For biomass Eq.( 3.18) was rearranged for least-squares analysis as follows:

28
(3.21)
(
where m is the number of biomass-time observations.

The derivatives of F2 with respect to A and kd give:

a:; = 2t. r11~ (A;:~:t') [(! -kd) - Y1 ln (A;:t) + ln (~)]


YSoXo (1- eat .) [(~ - kd) - qaln (A~!~:") + ln (t)]
(A + X o) (A + Xoe at .)

- ~' [C! -kd) - 1 y :t) +


ln ( A; ln ( ;- ) ]} (3.22)

F2=2~{(B_k)t2_YSotll (A+xoeat')+tl o)} (X (3.23)


kd f;; A dIA n A +X o 1 n X,

Equating thf' right-hand-side of Eq. ( 3.23) to zero and solving yields the estimate of

kd , denoted kd :
m m m
~t ln 1&1
+L-
Q1
2
ll.'\"t
A~ 1
-
A ~ 1 ln IAxQC
r.&'\"t AX0 , 1
X,1
kd = 1=1 1=1 m 1=1 (3.24 )
Lt~
1=1

Eqs. ( 3.22) and ( 3.24) do not correspond to the SSR of biomass concentration [i.e.

L:(X, -X)2 where X is the biomass concentration from Eq. ( 3.18)], but rather to the SSR

of the logarithms of the biomass concentrations [i.e. 2:(ln(X,) -ln(X))2]. By minimizing

the sum of square residuals of the logarithms (denoted SSRX), it is possible to simplify

the allalysis due to the fact that ka car be expressed as a function of A. By attempting

using 2:(X, - X)2 to estimate SSR, the values of kd and A would have to be estimated

sillluitancously by a numerical method as such as the secant method. This would make

l"
the equations and the numerical analysis more complex.

' '"
29
l.

3.1.3 Aigorithm for estimation of the parameters


1 Fig. 3.4 represents the algorithm developed in this study. The value of L* is varicd from

an initial (Lo) to a final (LJanaI) value with a given increment. To establish the f(,;l! iblc

region for L* (Lo < L* < L lanal ) it was observed that for values of L < Lo, le and ILm

were negativej therefqre, values of L lower than Lo were not considcred. The valut' of

LImai was established when' both SSR's tend to increase when L* increases. 1
The algorithm was cornputerized in a program, written in Turbo PASCAL v. 4.0 [6].

The prograrn creates an ASCII file, whe"eby for each value of L* the estimate of }l' is

found with Eq. ( 3.20). With the same value for L* and the estirnated value of r, the

estirnates of A and kd are calculated using Eqs. ( 3.22), and ( 3.24) (sec Fig. 3.4). Once

the pararneters are estimated, the surn of square residuals of substrate (dcnoted SSRS)

and SSRX are evaluated.

The secant method is formulated as follows:

(3.25 )

where f() is a function of x, Xn and Xn-l are initial estimates, and Xn+l is tht' ne\\'

estirnate.

3.1.4 Estimation of L

Finding an overall minimum by adding both sums of square residuals and plottilJg the

resulting surns vs L, was considered as being not an adequate technique. Thi5 i~ due tu

the faet that Eqs. ( 3.22) and ( 3.24) rninirnize the logarithm of the rnicroorganisfll~ LOB-

centration while Eq. (3.20) minimizes the soluble substrate concentratiOll. ConseqlJ('ntly,

rnostly the sum of square residuals of the substrate would contribute to d('temlll)(~ the

value of L.

1 Except for Run 4 COD, Run 5 COD and Run 5 TOC. See page 55

30
.J

1Begin 1
l
l IL* = Lo 1

Estimate Y by the secant method using Eq. ( 3.20) for


a given value of L*.

Once the value of Y is estimated, the value of A is found


by the secant mthod using Eq. ( 3.22). kd is evaluated
for each iteration of A using Eq. ( 3.24)

With the estimate of A, kd is estimated using Eq. ( 3.24)

After finding the values for all the parameters for a given
value of L*, SSRS and SSRX are calculated and saved in
the output file

[L* = L*+~LI
No
IIfL*=L jmal l
Yes
End

Figure 3.4: Algorithm

As an alternative, the SSR's divided by their respective minimum SSR were plotted

over the total range of values of L. In the following step, the minimum mtersection point

is locatedj this point is defined by the intersection of the two curves that corresponds to

the smallest m~~~R' This point represents the minimum percent that both curves are

"'off" from their optimum fit.

A comparison among the techniques available in the literature and the methodology
l proposcd hercin 15 shown in Table 3.1.

31
1

Table 3.1: Com2arison of methodologies


Reference Mathmatical Comments Paramctel's
technique esfzmated

Reynolds and Linearization Five or more batch reactors }r, kd


~
Yang (1966) are required.
: ~
;
~
Xo, Sa error-free
<
i-

Robinson and Gaussian Minimization of S, - St ILm, !\!, y


Tiedje (1983) X o, So error-free
Time as independent variable

Robinson and Gaussian Michaelis- Menten model Yrnax, !\'m


Characklis (1984) Time as independent variable So, R
-,
Simkins and Marquard Substrate as independent variable X;, So
Alexander (MARQ-FIT) Better fit with logistic model I!.m.
K.
( 1984) eventhough J(! was close to So

Braha and Linearization, Substrate as independent variable JLm, }/


Hafner Marquard X o, So error-free f{!, kd
(1987) Questionable methodology

This study Least-squares Time as independent variable


X o , So error-free
Assumption: f{5 > So
Non biodegradable
substrate

.i

f 32

1
3.2 Initial Estimates

To perform the secant method in Eqs.( 3.20) and ( 3.22) initiaIs estimates of Y and A are

required. Renee, to evaluate these initial estimates and to ensure that the roots obtained

were the optimum values, the following analyses were performed.

(a) Y from Substrate Plots

For a given value of L, ~~ [Eq. ( 3.20)] and SSRS were evaluated as a function of Y.

The results of these calculations are plotted in Fig. 3.5.

3.5 the initial estimate of Y ean be found precisely

*.
By looking at the plots of Fig.

either by finding at the graph the minimum value of SSRS, or the root of Also it can

be seen that only two roots are feasible for ~, one which minimizes the SSRS and the

other that maximizes it (when Y = 00).

(b) A from Biomass Plots

Once the value of Y has been estimated by the secant method through Eq. ( 3.20) for

the fixed L, Eq. ( 3.22) and SSRX were evaluated for different values of A. The plots are

shown in Fig. 3.6.

Looking at Fig. 3.6, it can be seen that only one value of A minimizes the SSRX

and two maximize it (A = a and A = 00, for both of which SSXR -+ 0 ). It is noted also

that. Eqs. ( 3.18) and ( 3.22) are discontinuous at A = O.

In the algorithm the value of L* was varied from Lo to L'anal. Therefore, initial

<'stimates of Y and A for each value of L* were required. Hence, the initial values for Y

{ and A for the Ilcxt L" were the roots obtained for the previous L*. Jt was found that when

the incremcIlt of L* was small (0.000001) these new estimates performed satisfactorily.

33
1

LI..
"0 o
y

CI}
cr
CI}
CI} 0

Figure 3.5: * and SSRS versus Y

34
Discontlnulty at A = 0

Dlscontlnulty
at A = 0

X
Il:
II)
II)
o A

Figure 3.6: !!fi and SSRX versus A

35
3.3 Laboratory Experiments

3.3.1 Sampling

The sample of sewage used in the analysis was a daily composite sample obtained from the

Granby (Quebec) wastewater treatment plant, which is an extended acration activated

sludge system.

A daily composite sample consisted in small samples taken every 15 minutes during

24 hrs in the influent of the basins. The daily samples were refrigerated at 4 oC un ti! the

last of the consecutive sampling days was reached.

The mixed liquor (called sludge hereafter) tested consisted in a grab sam pie frolll the

aeration basins. The composite sewage and the sludge sample were transported to the

Environmental Engineering Laboratory at McGill University immediately after the sludge ~

sample was collected. To ensure that the microorganisms had enough food. the ~Iudgc

(four to five liters) was fed with one liter of the composite sample of sewage, and acrat.ed

overnight at 4 oC in the laboratory. Batch reactor experiments stated the next 1ll0l1liIlg.

3.3.2 Batch reactor

For the initial population of microorganisms, approximately thrce to four liters of the

above sludge were left to settle for approxima.tely 30 minutes, thcn the supcrnatant was

removed. The settled sludge (0.5 to 1 L) was used in the batch rcactor expel illlcnt.

Four to five liters of influent were poured into the batch reactor togcther with the set-

tled sludge, and placed in an incubator (10 or 20 OC). Aeration and mixing were provided

to achieve complete mixing.

The analyses carried out at different time intervals during the expcriments (sec Table

4.10) were:

36
Mixed liquor volatile suspended solids (MLVSS) (triplicates),

Soluble chemical oxygen demand (COD) (triplicates),

Soluble total organic carbon (TOC) (triplicates),

Dissolved oxygen (DO).

MLVSS and COD (closed reflux) were performed following the outlines presented in

the Standard Methods (1976) [1]. TOC was measured using the Dorhman DC-80 auto-

mated laboratory analyzer (Dorhman, Santa Clara, California). DO measurements were

obtained using a Y.S.I. Model 54 Oxygen meter (Yellow Spri.lgs, Ohio).

3.3.3 Experiments

As commented in Chapter 2, two industries irregularly discharge their wastes into the
.\
sewage. Therefore, three preliminary runs were performed to determine the time required

by the microorganisms to degrade the biodegradable substrate.

For the preliminary runs the sewage used was a dail} composite sample of the previous

24 days. These runs lasted for 3, 24 and 30 hrs. From the last two runs, it was concluded

that ni ne hours was the approximate time required for the microoganisms to deplete the

biodegradable substrate.

37
Chapter 4

Results and Discussions

Estimation of the constants was carried out using three different data sets: (1) synthctic

data, (2) data from the study of Braha and Hafner (1987) [7], (3) data from five ex-

periments using wastewater and sludge from the wastewater treatment plant of Granby,

Quebec.

4.1 Synthetic Data

Initially the model used two sets of synthetic data for both substrate and biomas~. 1'wo

sugnificant figuref were used because this represents the accuracy of the me<tslIl'{'mcnt'i

performed in oU'. laboratory. The synthetic data were obtained by fixing thp value" of

the four pararl1eters in Eqs. (3.6) and ( 3.18) and assuming values of 50 and X Theil

the sub~trate and biomass concentrations were evaluated at diffel'ent times. Both the

synthetic data and the estimated constants are enclosed in Tables 4.1 and 4 2.

The estimated values of 1(, and Pm are 7.2 and 0.36 percent "off" of the original COIl-

stants, for SI and S2 respectively. Nonetheless, these new estimates yicld smallcr SSIt'~,

for both biomass and substrate, than the original parameters.

38
Table 4..
l' Synthetic data
Time (hr) SI S2
COD MLVSS COD MLVSS
0 100.00 1000.00 100.00 1500.00
0.25 94.57 1005.37 86.83 1505.33
0.5 89.40 1010.34 75.30 1509.00
0.75 84.48 1014.94 65.24 1511.21
1.00 79.80 1019.17 56.47 1512.13

1.25 75.34 1023.05 48.85 1511.93
1.50
. 71.11 1026.59 42.22 1510.74
1.75 67.10 1029.81 36.48 1508.70
2.00 63.29 1032.71 31.49 1505.91
2.25 59.67 1035.31 27.18 1502.49
2.50 56.25 1037.63 23.45 1448.52
2.75 53.01 1039.68 20.23 1494.08
3.00 49.94 1041.46 17.44 1489.24
3.25 47.04 1043.00 15.03 1484.06
3.50 44.30 1044.31 12.96 1478.58
3.75 41.71 1045.38 11.16 1472.86
4.00 39.26 1046.25 9.26 1466.94

Table 4.2: Original and estimated parameters for the synthetic data analysis
J(~ /lm y kd SSRS SSRX
( ~,) (hr-I) (mg biomas~) (hr-t)
m ~ub~trate
SI 3000.0 1.00 1.500 0.010 1.53 xl0- 4 1.44 X 10- 10
Estimated 1 3226.5 1.07 1.495 0.010 1.52 xlQ-4 1.06 X 10- 10
S2 4000.0 1.50 1.000 0.020 1.55 xlQ-4 5.16 X 10- 11
Estimated 2 4014.3 1.51 1.000 0.020 1.48 xl0- 4 4.08 X 10- 11

Figs. 4.1 and 4.2 show the feasible roots for Y and A. Looking at these plots the

same shape can be seen as those presented in Chapter 3, Figs. 3.5 and 3.6 (W vs Y
and ~ vs A). During the execution of the computer program, the values of Y and A

COIlvcl'ged, and SSRS and SSRX were minimized (Y, A, SSRS and SSRX were displayed

on the screen during the execution of the program).

39
S1
20

1 0
-20
-40
-so
-10
-100
-120
~ -140
...
~
-110
-110
-200
-220
-240
-260
-210
-lDO
0 2 1 1
Y

-1

-2
~
~,l,
.....
-3
,2-
~
.- E -.

-5

-6

-7

-1
-0.2 0 0.2 04 06 0.11
(Th .......... )
A

ay versus Y and ffi


Figure 4.1: Synthetic 1: .El aA versus A

4.2 Braha and Hafner (1987) Data Sets

In the study carried out by the above authors, seven experiments in batC'h reactors wer<'

performed to assess their methodology. Braha and Hafner l'an in parallel a complete

mixing, continuous-flow reactor for a period of approximately six months to providc ac

climated sludge for the batch experiments .


.... In the continuous-flow reactol', microorganisms were Ced with industrial W.l.stcwatcr

40

--------------- ------
52
10
'0
40
lO
20
10

-10
-20

~...
-lO
-40
-'0
" -10
-70
-80
-90
-100
-110
-120
-130
-140
2 4 1
Y

0.5

-05

-1

;n -1 :1
.!.
~2
....
..
f "
e
-2

-2.5

-l

-l.5

-4

-45
-200 0 200 400
,.

Figure 4.2: Synthetic 2: ffi


Y
versus y and BF2
BA
versus A

containing organic pollutants at different hydraulic retention times (4 - 17.5 h). For each

hydraulic retention time, the process was operated between 15 to 22 days to achieve

quasi steady-state (transitional periods 7-10 days not included). For each continuous-flow

experiment, samples were collected and a batch experiment was l'un on each of these

samples. The results of the seven batch reactors are shown in Table 4.3.

The methodology was applied to the seven data sets. The resulting graphs of the SSR

41
f; $.1 " ~ ...
Table 4.3: B!aha and Hafner (1987) Data Sets - ----

Expenment Analysis Time (hr)


0 0.25 0.5 1 2 3 4 5 6 7.5 10 12.5 15 24
LBI TOC !!!!l.
1 277 1 270 266 249 222 198 172 158 144 119 75 65 48 19
MLVSS T 5880 - - 570 - - - 5918 - - 6042 6075
- 6105
LB2 TOC !!!S.
1 281 272 264 249 221 197 174 154 141 119 79 61 46 20
MLVSS 7 4310 2 - - 4265 - - -4390 - 4481 - . - 4540 4596
LB3 TOC~
1 319 308 298 281 248 222 195 173 155 126 96 72 51 36
MLVSS 7 4190 - - 4148 - - -4319 - - 4412 - 4458 4488 .
LB4 TOC !!11l
1 296 286 278 261 233 207 184 161 142 121 88 60 49 33 1

AILVSS T 4020 - - 3970 - - -4128 - - 4246 - 4279 42931


LB5 TOC !!!S.
1
281 272 263 250 223 199 183 172 - - 92 58 5] 27 1
MLVSS~ 3680 - - 3641 - - -3756 - - 3861 - 3929 3948
LB6 TOC !!!S.
1
273 - 255 244 219 203 186 - 131 -118 84 50 42 19
~
l'\j
MLVSS~ 5020 - - 4929 - - - - - - 5150 - - 5237 5260
LB7 TOC~
1
228 220 217 197 178 158 119 97 85 77 63 55 38 -
AfLVSS 0/ 3880 - - 3821 - - - 3997 3 - - 4030
------ -----
4084
- - -
- -

Il 7 in the paper.
z430t3 ln the paper
34097 in the paper, beheved t be 3997.
ratio, and of the fitted substrate degradation and biomass growth curves (both obtained

with the parameter estimates) appear in Figs. 4.3 to 4.9. Table 4.4 shows the feasible

region of L and the estimates obtained.

In the computer program, the value of kd was constrained to be non-negative in any

iteratiol1. To constrain the value of kd , a value of zero is assigned to it if it is lower than

zero from Eq. ( 3.24) 'during the search procedure for A (see algorithm Fig. 3.4).

Table 4.4: Estimated parameters with Braha Lnd Hafner (1987) data sets
Expe7'iment Lmm Lmax K5 J.tm y kd L
1
mQ hr
1
mq hr
!!:.S.
1
hr- J m!lMLVSS
m~Q TOC
hr- 1 1
m~Q hr

LBi 0.000015 0.00020 297.53 0.0068 1.192 o+ 0.000023


LB2 0.000026 0.00010 4868.41 0.1266 0.958 ot 0.000026 t
LB3 0.000027 0.00030 216127.90 5.836 0.961 ot 0.00002i t
LB4 0.000029 0.00016 12777.22 0.3705 0.993 ot 0.000029 t
LBS 0.000029 0.00038 3215.28 0.0965 1.017 ot 0.000030
LB6 0.000020 0.00100 36.61 0.0037 4.805 0.0003 0.000100
LB7 0.000033 0.00073 12041.24 0.3974 0.922 ot 0.000033 t
t L = Lman t Constraint WOlS active

The constants obtained are shown in Table 4.4; the results vary greatly. Eventhough

the sludp:c age varies among the experiments, this could not account for the differences

in the parameters estimated. On the other hand, the shape of substrate degradation

and biomass growth curves are similar for the seven batch experiments and should yield

similar values for each of the parameters.

It IS notcd in six (6) out of the seven (7) SSR ratio graphs (Figs. 4.3, and 4.5 to 4.9)

that the m~:;Y:U plots reach a plateau with shallow slope at certain value of L, denoted

here as Lp. For values of L < Lp, the constraint on kIJ was active (i.e. kd < 0 -+ kd = 0),
while for L > Lpl kd was > O. For LB2 the value of k i remained constrained throughout

the fClslblc Icgioll of L.

Silllil,\! to the s)"nthetic data analysis, the partial derivatives of FI and F2 VCI sus }~

43
lB1
1.1:1
~.
1.14
;+
;"
.+
1.13
'r

i
1.12

c
1.11

~ .+
1.1

~ 101

III
4
1.011
107
~ +
tl 1.01S

f 1.D!5
+
1.04
R
1.03
1.02
+
1.01
1
000001 O.OOOOl 0.0000' 000007 000009 000011

L 810 ...... _
Subatnlt. +

Biomass
&.12
&.1
&.08
&.05
&.04
&02
li
'.98
~~
"'"
EC:
~::
U5
'.94
en"
~
%
592
'9
SB8
SB6
'.84
, 82
'8
'.78
0
'.75
0 12 16 24

trno (hr)

Substrate
280

250
240

220
200
180
Q
160
~
! 140
8... 120
100

80
50
40

20

0
0 , 12 \6 20 24

trno (hr)

Figure 4.3: SSR ratio, biomass and substrate graphs for LBl

44
'J

LB2
1.5

l ..,
i
c:
~
1.4

~
t!
4 I.l
~
la
i 1.2
~
1 1.1

0.00004 0.00006 O.OOOOB 00001


0.00002
L
- Substrat. + 1........

Biomass
4.5
4.'8
4."
4'4
4.'2
4.5
4.41

., ."
~-;'
ec:
'"'l:
4.46
444

~l
442

.... 4.4
%
411
4.l6
434
412
4.l
4.28
4.26
0 12 \6 24

tme (tir)

$ubstrate
lOO
280
260
240
220
200
110
~
.s 160
140
~ 120
100
la
60
40
20
0
,
'1 0 1 12 \6 24

~ t.rn. (tir)

Figure 4.4: SSR ratio, biomass and substrate graphs for LB2

45
'J
LB6
1.5....-----------------l

1.1

0.0003 000034
00001 000014 0 DOOn 0.00022 0.00026
000002 000006

- - Substrat.

Biomoss
5.l-r-----------------,
525

52

4.95
D

'9J---,---,----r--~a~-,--~12---.---:.,6~-.---:~-.~~2.
o
lino (h.)

Substrote
2ao~----------------------------------------~
25C
240

220
200
IBO
~ 160
.g 140

~ 120
100
aD
&0
40

20

a 12 16 20 24
o 4
'J. lino (IIr)

Figure 4.8: SSR ratio, biornass and substrate graphs for LB6

49
.j

1.7
LB7
l
\.5

ic 1.5

~
~ \.4
t!
4
I! \.3
~
l 1.2
:.:r
i
\.\

1
0.00002 000006 0.0001 000014 000018 000022 000026 00003 000034
L
- Su~tnrte + 8iG....

Biomass
4.1

4.08
4.01
4.04

4.02

~';'
..... 3.98
e c
"'l: 3.96
111"
~~
::lE
J.94

3.92
3.9
J 88

J.86
J.84

3.B2
a 2 4 10 12 14

lft-.. (h')

Substrate
210
220
210
200
190
IBO
170
160
150
~ 140
5 IJO
8
~
120 c
110
100
c
90
Ba o

--
70
60 c
______ c
50
40
la
a 5 la 12
lft-.. (h')

Figure 4.9: SSR ratio, biomass and substrate graphs for LJ37

50
and A respectively (see Figs. 3.5 and 3.6) were evaluated at the minimum intersection
( point (i.e. lowest intersection point in the SSR ratio graph)j the shapes of the partial

derivat.ive.'> show a single root. The plots are presented in Appendix F.

Table 4.5 shows the original constants estimated by Braha and Hafner (1987) [7].

The table show the SSRS with the parameters found in the original study, and the SSRS

obtained by the methodology used herein. It can he seen that the sum, of squares ohtained

in this stl:dy are smaller than those given by the original study.

Table 4.5: Parameters estimated hy Braha and Hafner (1987)


Experiment Y J(~ Pm 4 SSRS SSRS SSRX
ms. MLVSS !!!S. hr- 1
Braha and This study 10- 3
mg TOC 1
Hafner (1987) [7]
LBl 1045 14000 00400 223.2 220.7 2.50
LB2 1.35 13000 00482 103.4 93.8 5.17
LB3 1.37 13000 0.500 509.4 442.7 7.11
LB4 1.38 14000 0.568 384.1 352.7 7.13
LBS 1.56 15000 0.776 1096.4 341.1 7.26
LB6 1.42 16000 0.548 725.3 413.8 3.66
LB7 0.78 14800 0.383 1490.1 1125.9 3.41

4.3 Data from the Granby Plant

The Granby wastewater treatment plant treats the sewage from the municipality of

Gran by, Quebec. It is known that two industries (cheese processing and textile) dis-

charge their waste irregularly into the city sewage. The organic load from the cheese

processing industry is approximately one third of the total organic load of the sewage.

The load from that industry is mostly in the form of organic suspended solids.

The wastpwater treatment plant consists of an activated sludge pro cess that is op-

t'rated in the extended aeration phase. Theoretically, th,~ extended aeration process is

4The Ulllts 10 the orrgmai study are d- 1 , but upon our calculations the units of Jl.m are hr- 1

51
designed such that aU substrate removed is converted into energy metabolism and oxi-

1 dized, so no ex cess biomass is produced and sludge handling is eliminated. Hencc, the

rate of growth can be written

(".1 )
.
In other words, aIl the biomass produced is theoretically oxidized and metabolizcd by

microorganisms. This assumption was proved by Gaudy et al. (1970 [12], 1971 [lI}, 1974

[35]), and Obrayashi and G \.udy (1973) [22].


The plant consists of six stages as indicated in Table 4.6.

Table 4.6: Characteristics of the Granby wastewater treatment plant


Stages , Pumping statwn
, Sand removal
, Screening
, Aeration basms
, Secondary clanfier."
Ch/orinatwn
Equivalent population 126000
3
Inflow average 50000 dBy
m

Aeration basins (4) 25693 m3 total


Average hydraulic retention time 12.3 hr
Sludge age 40 days

4.3.1 Experiments

Initially, three preliminary experiments were performed at 20C in order to monitor the

behaviour of biomass growth and substrate degl'adation. From thc!>e exrcriment~ th('

presence of non-biodegradable substrate in the sewage was obscrvcd. lt is bc]ieved that

most of the concentration of non-biodegradable substrate cmanated from the textile ill-

dustry. Also, it was noticed that thp. degradation of the substrate was achicved within

nine hours or less. Nevertheless, substrate measurements wcre obtaincd in Runs l, 2 and

52
.5 for 2.5, 24 and 48 hours after initiation of the tests, respectively, but the concentration

levels of substrate did not decrease after the tenth hour.

The (ive experiments were performed in the following dates: June 28, 1989, August

23, 1989, December l, 1989, February 3, 1990 and February 10, 1990. Unfortunately, the

conditions of the biomass in the plant changed considerably from experiment to another.

In the last week ~f July 1989 and first three weeks of August, major repairs \Vere

perfOl med in the apration system of the basins. During that period, once a week, . dif-

fcrent tank was emptled for such repairs. This forced the biomass to adapt to the new

condItions, which were different than those in the first experiment.

During the second week of November 1989, sludge bulking occured in the plant.

Bulking is the result of preferential growth of filamentous baeteria among the various

(non-filamentous) speeies of microorganisms normally present in the aeration tanks. The

filamcnlolls bacteria usually hinder the settling of flocs in the seeondary clarifier and, con-
..
sequently, large amounts of biomass is lost in the effluent. To extinguish the filamentous

bactel iJ, the basins were chlorinated during the fourth week of November. As indicated

previously, the third experiment was started in December 1, 1989 using samples (i.e. chlo-

1 inatco) from the fourth week of November.

The last two experiments were similar between them, but climatologie conditions (i.e.

much lower operating temperature) were considerably different than those in the first

th rel' batch 1 li n!>

The ,>ewag<. tcsted presented a certain amount of non biodegradable substrate, and the

mode! llscd ;n this study (Monod kinetics) consid~rs only the biodegradable substrate.

lIenec a I1Icthodology is developed to assess the amount of non-biodegradable substrate

ill tlll' sdll1ples.

The ITldhodology consits in giving different concentrations of non-biodegradable sub-

..,trelte ,\Ild thpll subtraeting this amount from the total substrate (measuremcnts). Then

53
the sum of squared residuals is obtained following a negative exponential trcnd mode!.

The model proposed is given by the following equation:

where S, is the biod~gradable substrate at time ti (i.e. total substrate mlJlus nOIl-

biodegradable) and bis a constant. Rearranging for least-square analysis it yiclds:

n
Fo(b) = ~)Soe-bt, - S,)2 (<1 :1)
1=1

Performing the derivatives with respect to band rearranging gives:

F
_0
b = 2S 0
{n0
~t
S L- ' e- 2bt , - n S e- }
~t
L..J 1 e
bt ,
(1.4 )
1=1 1=1

where bis estimated by the secant method when the derivative in Eq. ( 4 4) is set to be

equal to zero.

The Ilon-biodegradable substrate estimate is varied over a wide range. The mo(l!1

yields one minimum SSRS for each substrate-time data set, which was con"'ld(>ICU tu 1)('

the non-biodegradable substrate, although this concentration did not corr('.,polld to II\('

last substrate concentration measured. The computer program is glvcn in Appcndix ('

The sludge age of the biomass on the last day of sampling, the temperature at which

the experilY'~nts were carried, and the nonsoluble substrate ale pre!:>cntcd III Tdb!c .,

Results of the five experiments are presented in Table 4.10.

Table 4 ..
7' Characteristics of sludge and substrate for each experiment
Run Sludge Age AfLVSS SVI F/M Temperature N on-bzod cgmdablc.
days !!13.

!E.'l
J
day-l oC COD (T) TOC (7)
1 35 - - - 10 44.31 :n O
2 46 3007 57 - 20 63 ..58 2808
3 48 2300 59 0.10 20 98.00 ~2.20
4 43 2316 71 0.11 10 123 87 19.24
.5 40 2316 71 0.11 10 132.07 .56.DG

54
Contrary to the analysis with the data sets from Braha and Hafner (1987) [71 paper,

the value of kd was always positive, therefore no constraint was needed.

On the other hand, L fmal was constrained to values su ch that the corresponding l(~

were larger than 12. This constraint was considered due to the fact that large values

fol' A's <tIC cxpf'cted (Simkins and Alexander (1984) [28], Braha and Hafner (1987) [7],

Uloth alld Manavic (1977) [32] and Lawrence and McCarty (1970) [17]). The constants

cstillliLLcd and the plots for each data set are shown in Table 4.8 and Figs. 4.10 to 4.19,

respectively.

Table 4.8: Estimated parameters for the Granby Plant data sets
Run Lmm Lmar J(~ J1.m y kd L
1
1
mq hr
1
mq hr
!!!S.
1
hr- 1 m2\fLVSS
ma Substrate
hr- l
mq .'Ir
1 COD 0.000149 0000395 637.0 0.103 0.268 0.002 0.000161
TOC 0.000423 0.001930 1313.6 0.552 1.092 0.002 0.000423 t
2 COD 0.000141 0.000540 2783.0 0.392 0.416 0.00.5 0.000141 t
TOC 0.001410 0.006310 3464.1 ~.884 1.284 0.006 0.001410 t
-
:3 COD 0000:312 0.000761 883491.0 275.650 0.606 0.009 0.000312 t
TOC 0002490 0009490 2757.8 6.867 4.577 0.011 0.002490 t
.,.
4 COD 0.000082 0.000422 338.3 0.032 0.081 0.002 0.000094 +
TOC 0000280 , 009380 1802.4 0.505 0.704 0.00:3 G 000280 t
5 COD 0.00011 0 0.003240 14.1 0.006 0.395 0.002 0.000420 ~
+
TOC 0000800 J.004100 155.1 0.132 0.583 0.001 0.000850 +

t L = Lmm + Constraint was active

Tahle 4.9: Ratio of soluble to total substrate concentration (COD) at Granby


m~an standard devzatzon number of analysls
Influent 0.3?14 0.0646 10
Eillucnt 0.7439 0.1284 1.5

Analyzing the ploU, of runs 4 and 5 (COD), it can be seen that the phenomenon

of contact stabilization may have taken place in our batch experiments considering that

f Illost of the mcasurcmcnts after the 6 th hour are above the substrate predicted and the
1

'1
1

.55 1

j
- ----- 0 _ _ i Ji , uza'

s-";
....

Tahle 4.10: Snb.,t rate and biomass data for runs 1 to 5


Run 1 Run 2 Run 3 Run ..( Run 5
Time COD TOC MLVSS COD TOC MLVSS COD TOC MLVSS COD TOC MLVSS COD TOC MLVSS
/zr ~ ~ ~ ~ ~ ~ !!.!9. !!.!9. !!.!9. ~ !!.!9. !!!1l. !!!1l. !!!1l. !!!1l.
T T T T T T T T T T T T T T T
0 121.0 46.4 1830 160.0 62.27 1421 212.5 39.93 651 271.0 73.74 770 242.3 75.13 909
0.5 73.0 37.Cl 1818 120.8 - 1156 180.0 4?.l 702 237.0 - 771 209.4 76.3 918
1 68.0 34.1 1857 120.0 33.97 1510 171.0 30.2 685 205.0 69.8 781 - - -
1.5 69.0 30.9 18n 125.0 31.18 1406 151.5 32.0 671 146.3 54.44 776 165.2 50.01 926
2 63.0 30.7 1862 100.0 31.78 1429 156.5 29.l 670 133.5 55.83 780 135.2 50.48 910
3 - - - 80.0 30.55 1468 145.0 28.8 6SS 126.5 61.24 772 128.4 58.36 908
-1 - - - 69.0 27.67 1427 133.0 25.7 &65 115.1 70.1 766 121.3 50.63 912
CJ1
Cl 4.5 42.0 24.3 1812 - .- - - - - - - - - - -

.5 - - .- 83.0 29.5 140tl 130.0 29.3 666 - - - - - -


5.5 - -- - - - - - - 110.4 53.94 772 135.8 61.12 918
6 42.0 23 ..5 1839 - - - - - - - - - - -
7 - - - 70.9 28.9 1:388 104.0 21.9 660 134.0 48.53 768 145.0 58.34 907
S - - - - - 970 23.7 679 - - - - - -
$5 - - .. .. - 132.1 49.04 2 126.6 53.22 904
9 - - - 9q ~ 21..1 666 - - - - - -
10 10.0 22.2 lK12 58.0 2.3.12 lt 10 - - 112.0 51.00 no 139.2 60.67 916
11.5 . - - . - - 128.8 49 10 765 137.9 .59.23 904 :
RUN1
1 DOS

10011

l i
C
1.007

~~
1006

1 DO!!
4
&1 1004
t1
!l 100J
E

~ 1002

1.001

000014 o OOOIS 000022 0.00026 o OOCl 000034 000038

Subatnrte
L
.. !hon.o

Biomoss
1.9
189
1 Sil
I.S7
186 C
0
1 S5

~~ 1 S4
"'''
EC 1.83
v~
1/1" I.S2
c
~~
C
::E I.SI
'l'
1S

'. 179
17S
177
1 76
1 75
0 2 4 ID
t..... (hr)

Substrote
130

120

110

100

90

~
BD

.s
0
70
0
0 60

50
0 0
40

3D

20

ID
0 2 4 ID

~ liN (hr)

" Figure 4.10: SSR ratio, biomass and substl'ate graphe;; for l'un 1 (COD)

57
Figure 4.11: SSR ratio, biomass and substrate graphs for run 1 (TOC)

58
RUN 2
108
COD

1 07

ic
1 06

~ 1.05
~
~
~ 1.04
1/1
~

~
1 Dl
~
;;;. 1 02
~
101

1
00001 00002 o OOOl 00004 00005
L
Subolnll. + 810"",.

Biomass
1 ~6

1 ~4

1.~2

15

148

,,...,........ 1.46 0
0

e c:
~::
1"
"'~ 1 42
~. ~
::t 0 0
1.4
0
1 lB

1 l6

134

1 l2

13
8 la
!me Chr)

Substrate
180
170
150
.,0
140
IJO
120
~
.s
Q
110
100
0
V 90
80
70 0 0

60
~o

40
JO
6 la
tme (hr)

Figure 4.12: SSR ratio, biomass and substrate graphs for run 2 (COD)

59
RUN J
1.22

-1
121
1.2
1.19
)(
1.la

~
1 17
1 16
c 1.15
{ 1.14
~ 1.13
~
. 1 12
1.11
Il
X1
1.09
tl 1 oa
~ 1 07
;;;. 1 06

~ 1 05
1 04
1 03
1 02
1 DI
1
00006 o ODOS
00003 00004 0.0005

L
- Subtltnrte + 810""'.

Biomoss
740

730

no
710

700
<:>
"-
co
690 0
0
4' t 6S0
0

1 ~
~
670
0 C
0
660

650

640

630

620
0
tme (hr)

Substrote
240
2JO
220
210
200
190
ISO
,.
<> 170 0

g 160
150 0
Q
0
0 140
130
120
110
0
100 0
90
SO
70

t 0
trne
4
(hr)
6 a

Figure 4.14: SSR ratio, biomass and substrate gl'aphs for l'un 3 (COD)

61
RUN 4
COD
I.J'
I.J2

( I.J
1.28

i 1 25
12'
"
~ 1.22
~ 1~
t'! 1.18
4
fil Ils
0:
t'! 1.1'
1 12
"
;,. Il

~
1 08
1 06
la'
1.02
1
000005 000015 000025 0.000J5 0000"
L 81a .... _
- Substrole +
Biomoss
790

785

a 0
780

C
<> 775
"- c c

~
.s'"
fil
770

il ~ c
:1
765

760

755

750
5 la 12

trM Chr)

Substrote
300

2BO

260

2~

220

~ 200
.s
Cl
0
IBO
U
160

1~ a
c c c
C
120
0
C
100

f
80
a , 5 8 10 12

" \me Chr)

Figure .t.16: SSR ratio, biomass and substrate graphs fol' l'Un 4 (COD)

63
RUN 4
TOC
1.22

1 12

I.IB

i
c
1.15

~ 1.14
~

..tI 1.12

~ Il
tI
lOB
~
~ 106
~ 1.04

1 02

0 0002 0004 0006 o OOB 001


L
- Subslnlte + 810""'_
Biomoss
790

7B5

C
780

775
~
... C
!
... 770

~
:E C ~~
765

760

755

750
0 2 4 B 10 12

lrne Ch')

$ubstrote
100

90

BD

70

~
! 60
g
1-
50 C C

40

JO

~ 20
0 2 4 6 B 10 12
...t trr. Ch')

FigUl'e 4.17: SSR ratio, biomass and substrate graphs fol' l'un <1 (TOC)

64
RUN 5
COD
122

l 1.2

1.111

1
c
1.11

1.14
itI 1.12
~ Il
t'I
c 1.011

~
1.06

1.04

1.02

0 0.0004 00008 00012 0.0016 0.002 0.0024 000211 o 00J2

L
Substnlte + liII""'.

Biomoss
~Jo

c
925

920
C C
C
915
~
!'"
C
-d
910

~
::E
a
905
D D

900

1195

1190
0 2 4 5 ID 12
ti11e (hr)

Substrate
2So
270
260
250
240
2JO
220
210
~ 200
! 190
Q
0 1110
u
170
160
150
a
140 a a
a a
IJO a 0
120 0
110
100

'", 0 2 4
ti11e (hr)
5 1 10 12

Figure 1.18: SSR ratio, biomass and substrate graphs for run 5 (COD)

65
RUN 5
TOC
1.17

1
1.111
1.115
1.14

1 <:
1.13
1.12

~~
1.11
1.1

..
I!
1.01
1.01

~ 1.07
<: 1.06
E
~ 1.015
i~
1.04
1.03
1.02
1.01
1
0.00011 00012 00016 0002 o OOH 000211 00032 00036 000'
L
- - Substrnle + 1110 ......

Biomoss
930

D
925

120
D D

..
~
E
'-'
Sl5

910 a
~
%
905
0 0

900

B95

B90
0 4 10 12
tme Ch')

Substrote
100

90

110
a
70
Q

~
! 60
D
C
0
C
0
8
1- D
50 o D 0

40

JO

20
2 10 12
tme (h,)

Figure 4.19: SSR ratio, biomass and substrate graphs for l'un 5 ('fOC)

66
Table 4.11: SSR's obtained with the Gr anby Plant data sets
l Run SSRS SS RX
10- 4
"
1 COD 483.81 5 .93
TOC 8.54 5 .97
2 COD 839.53 20.1
TOC 37.90 3 9.2
. 3 COD 624.34 24.9
. TOC 72.52 6 4.9
4 COD 2251.76 3.0
TOC 366.99 3.7
5 COD 839.19 5.2
TOC 335.14 4.4

fc\\' Illcsurements prior to the 6th hour are below the substrate predicted. Ratios of sol-

uble t total sllbstrate concentration (COD) were obtained from tests performed by the

persone! of the Gran by wastewater treatment plant during the months of November 1989

clnd ,Jcl/luary 1990. The results are presented on Table 49 which indicate that contact
''1.

slctbilization was probably taking place in the Granby plant.

l\Clthcr the differences of the characteristics of sludge and substrate in each experi-

l11Cllts liaI the contact ~tabilization phenomena could account for the values of the pa-

ranll'ters c5tlfnated in this study. The parameters are far from being simlilar, except kd

(Ind } (in SOITlC cases). It is known also that the differences in tempe rature during the

pel formancc of the experiments affect the parameters estimated; however it is Bot possible

ta e\'lluate prccisely such effects.

~~.! vs } - and ~/~; vs A \Vere evaluated for each run and only one foot was found in

each UISC (the plots are shown in Appendix F). As in the previous analyses, Y, A, SS' R.','

and SS'RX werc di5played on the screen of the monitor and it was observed that while V

(Illd A l'llll\Clged, SS RS' and SSRX \Vere minimized. Substrate degradation and biomass

gJl)\\,th were weil fitted in ail the plots, Table 4.11 shows the SSR's obtained for these

dat a sets.

67
In summary, the analyses carried out using data sets from Braha and Hafner (198i) [il

1 and the Granby Plant have yielderl values for 1<3 and !Lm that are far flOm bt'inp; simil.1I

betwecn them. Nevertheless, the estimated values for Y and kd were snnilal lnd \\'Ithlll

the values commonly found in the literature. The main concern rcmaillS in 't.he l'!>timdtioll

of /{s andwhich seems to be a very difficult task due to the high H!e!atiun bl'l\\"('('ll
Jlm

these two constants [see Robinson and Tiedje (1983) [27] and :\ppcndlx A].

On the other hand it is known that the minimum sun! of square resldllals is t 1)('

criterion which has to be satisfied in numerical approaches for kinetic est; rllation III t I\('

chemical and environmental fields. This study has shown the missesttnmtioll ln pn'\'IOIIS

studies of the Monod kinetic parameters which did not mect thi.., niterIOll [BI(lha alld

Haller (198/) [7] and Simkins and Alexander (1984) [28] (~lonod \\ith g\O\\th)] The ".,-

sessment developed herein has shown (Appendix F) that tlw 1lIl1limum SSH 's wa" Illet ill

every data ~,et analyzed. Unfortunatcly, the estimates of A. and !lm WCH' Hot (,Oll!->Istant.

and several times out of the values commonly found in the literature.

I
68
Chapter 5

Summary & Conclusions

To properly design a wastewater treatment unit, it is recommended to knO\v the

l'ales of su bslratc: uptake and biomass growth. The best known model to Pledict

these rdles IS the Monod kinetic model (which contains four parameters). The

estimation of Monod kinetics can be carried out either by continuous flow 01 batch

reactors. It is known that batch reactors is the best alternative in terms of time

needed to pel form the experiments .

It was lemarked herein that the parameters estimated in the studies by Braha and

Hafner (1987) [7J and Simkins and Alexander (1984) [28J (Monod with growth) did
not conespond to the minImum SSRS. Therefore there IS a need to develop a ne\\'

approach, to estimate the four parameters from batch reactor data sets, that meets

th", minimum SSR cnterion. Optimum estimates of parameters are obtained whell

this critcrion is met .

The Ilonlinear least-squares methodology presented herein, fits substrate utilization

and biomass growth curves simultaneously. The secant method was used to solve

the nonlinear Icast-squares equations.

69
The methodology developed in this study \Vas applied to three diffclent dat.a sets:

(1) synthetic data, (2) data from the study of 8raha and Hafner (19S7) [7], (3) data
1
from five experiments performed with wastewater and sludgc fro111 t.he Wd"t('\\',tI ('1

treatment plant of Granby, Quebec. The values found of f{$ dIld JI", wlth delta

sets (2) and (3) were dissimilar among them, and sorne of thcm are Ilot wit.hiIl tl\('

values cornmonly found in tht literature. lt is believed that the 1l1l(OIlCl'it.l'IHy of

the estirnated values for these two parameters is due to the high COI relat iOll bl'l.\\'('('11

them.

This assessment appears to be the first to estimate the four Monod-klllctic p,t!,lIn-

eters from a single nonlinear least-squares formulation. by fittmg silllltlL-IIl('oll"ly

substrate uptake and biomass growth.

Based on the criterion that species of mieroorganisms depletmg slrllllar ~lIb . . tl(ttl'

under similar envl;onmental conditions (pH, oxygen, tempclature, Lee. ) ~h()\Ild

yield about the same values for the parameters, it appears to he qucst.ionabk t.u

use this ne\\' methodology for the estimation of the four MOIlod-kinctlc!'> plI illlH't('I"

before more studies are performcd.

The minimum SSR is a eriterion tnat needs to be fulfilled to obtain the optlllllllll

estimates, which has been achleved with this new methodology

However, obtainmg accu rate estimates of the parameter5 seelll~ t he ct VCly cldli( 1Iit

task given that the response surface of the SSICs functioll scems to be quite fiat III

the vicinity of the minimum SSR point, as shown in Appendix A.

The parameters obtained in this research work cannot be cornpalf'd directly with

the values estimated in previous studies [e.g. Braha and IIafner (1087) [7JJ <'IIIU' \V('

have shown that these values do not fulfill the critcrJon of minimum SSR

70
Further studies need to be done in this topie, for example by running in parallel

1 continuous-flow and batch experiments with synthetic waste and comparing the

constants estimated in both systems. Also, different numerical approaches need to

bc attempted that could possible handle the high correlation among the constants.

71
1

Appendix A

Sensitivity Analysis

The analysis presented herein shows the existence of several sets of parlllleters whl( h

yielded smaller SSR than the parameters estimated by Braha and Hafllel (19R7) [] ill

thelr study.

A.l Introduction

This study consists in a sensitivity analysis which gives different values to each cou..,t,LIIt

and then in calculating the corresponding sum of squares residuals (SSR) wlth resp('ct to

the substrate, which is defined as:

n
S'S'R = 2)St - SI)2 (A 1)
1=1

where

s, = measured substrate concentration,

St = calculated substrate concentration, as obtained with Eq. ( 221)

72

1
,-
A.2 Model and Application

SSRS were evaluated by giving an initial and a final value for fJ-m, Y and J(! with their

respective increments. The computer program calculated the SSRS for each possible

cOlllbination of variables .
.
Duc to the [act that a value of St is reguired in Eg. (A.1) and since Eg. (2 ~1)

canllot be written in ter ms of substrate as a function of time, the St value was found via

the ~cCtnt mcthod. The InItiaI guess value was the corresponding S,.

Braha and Hafner (1987) [7] proposed a methodology in which the yalues of Y and

kd were round USIng Eg. (2.18) and the values of /{! and fJ-m were estimated through

Ilollltllear lea!>t-squdfes analysis with Eg (2.21).

l1egardlcfls of wbat these authors stated, and analyzing Table 2 in their paper, it seems

that the value of Y wa~ also estimated simultaneously with Ks and fJ-,n vIa nonlinear least-

'>quarf'S lnalysls, alld not usmg Eq,( 2 18) to estimate Y a~ It was established.

Due to the fact tbat we could not assure what methodology was applied, two analyses

\\"CI e pcrformed The first consisted In varying the value Y and the second used the value

obtailleu \Vith Eq. ( 2.18).

Inltially a SSRS \Vas computed using the parameters estimated by Braha and Hafner

III their study \Vith the first ex periment (f{s = 14000 mq~OD, fJ-m = 0.4 hr- I and Y =

1 15 Hl7~,'~~~~~~/I) The SSRS [('und \Vas 223.188; the proposed inItial and final values for

crlch of the pal amctcrs and their increments are presented In Table A.l. The outcome

Tahle A.1: Parameters studied


parameter initial value final value increment
/(3 100 20000 100
}" 0.5 4.1 0.1
Ilm 0.1 2.1 0.1

of this lllalysis gave 855 different combinations that yield sm aller SSRS than the one

73
proposed by Braha and Hafner (1987) [7]. The SSRs vary from 223.187 to 207.585. The

l values of the parameters that yield the thirteen smallest SSRSs are presentl-d III '1'<11>1(,

A.2.

Table A :~: Thirteen smallest values of the SSRs for tlH' first analysis
y [{~ J.lm SSR li.....
J'rn
2.8 9300 05 207.748 1~600
2.8 11200 0.6 207.791 1S667
2.9 10800 OC 207.707 18000
2.9 16300 0.9 207.702 1S111
3.0 14000 O.S 207.621 17500
3.0 19300 1 1 207 795 17515
3.1 15300 0.9 207.620 17000
3.1 17000 1.0 207 .5~5 17000
3.1 18700 1.1 207 721 17000
3.2 16500 1.0 2075SB 16.500
32 19800 1.2 207 :?6 16,500
3.3 14400 09 207.727 16000
3.3 16000 1 0 207.702 16000
3.4 18700 1 2 207.684 1,1.183

In the second analysis the value of}" was set to 1.'1.5 (since this was thc vetlue c~tlll\at('rI

for Y by the authors) and J{~ and J.lm \Vere varied from the same inltlLl to filla l val W'""

with the same increment as in the first analysis. Fourtecn differcnt combill,iolls )'ll'ld

lower SSRS values. The output is shown in Table A 3.

A.3 Comm~nts

The results obtained from this analysis show that estimates that ylcld SSRS lowcl, thall

those obtained by the mdhod of Braha and Hafner (1 9S) [7J.

These authors used the computer prograrn BP7 1 B with the dpplication module CIII v(!

fit They suge~ted that a nonlinear least-squales analysis wotlld be a reasollable tcchniqlle

to use. The algorithm in this computer program is Ilot known, but it scerns that the fitting

74
Table A..3' SSRSs values for the second analysis
l y J(~ J1.m SSR lia.
u_

1.45 3400 0.1 208.260 34000


1.45 6900 0.2 222.197 34500
1.45 7000 0.2 211.222 35000
1.45 10500 0.3 216.358 35000
1.45 10600 03 213.263 35333
1.45 10700 0.3 222.701 35667
1:45 14100 0.4 215 240 35250
1 45 U200 04 214 477 35500
1.45 14300 04 220.744 35750
1.45 17600 0.5 219.575 35200
1.45 17700 0.5 215.135 35400
1.45 17800 0.5 215 266 35600
1 45 17900 0.5 219.888 35800

technique employed nonlinear least-squares analysis.

After performing the SdIllC analysls with different initial and final values \Vith different

illClcments, it is observed that there is no pattern to be followed (between the variables

and the SSR) if one desires to estimate the parameters with the tnal-and-enol technique

It I~ unknowll if the computer Pdckage used in the original study estimates the pa-

1 dllletCI ~ by considenng eithcr the time or the substrate as the independent \'aI iable,

thcrefore, the same analysis was carried on but obtaining the SSR \VIth respect to time

(i.e substrate as independent variable). Varying Y, ](3 and /lm with the same initial and

filial values and inclement'i (similar ta the first analysis), 3369 sets of vanables yielded a

"'I\l,dler SSR.

75
Appendix, B

Main Computer Program

The algorithrn presented in Chapter 3, Fig. 3.4, is cornputirized in the program showlI

in thlS Appendix.

76
pragram Algorithm;
var infile,outfile : text;
tx,tt,ss,xx: array[O .. 20) of real;
srs,5rx,suml,sum2,5um3,sum4,sum5,sum6,sum7,sum8,5um9,nosol : real;
A,B,C,Xo,An,llinc,llfinal,lln,al,Yn,Y,So,Ks,mu,kd,t,s,x : real;
i,n,m : integer;

procedure captura;
begin
assign(infile,'c:\input.txt');
assign(outfile,'c:\output.txt');
reset (infile) ;
n:=O;
m:=Q;
readln(infile,n,m);
for i := 0 ta n do
begin
readln(infile,t,s);
tt[i):=t;
55 [ i) : =s;
end;
for i := 0 to m do
begin
readln(infile,t,x) ;
tx [i) : =t;
xx [i) : =x;
end;
close (infile) ;
rewrite(outfile);
end;

function ssrs(Y:real):real;
begin
suml := 0;
for i:= 1 to n do
suml := suml + 5qr(y*so+Xo)*So)/(Y*So+Xo*exp(11n*tt(i)*(So+Xo/Y)-ss[i);
ssrs . - suml;
end;

function ssrx(kd,A:real):real;
begin
suml : = 0;
for i:=l to n do
suml := suml + sqr(ln(Xo*expB/A-kd)*tx[i]-C/A*lnA+Xo*exp(al*tx[i]/(A+Xo
ssrx .- suml;
end;

function real) :real;


var den : rL _,
begin
suml:=O; sum2:=O; sum3:=Q; sum4:=0;
for i := 1 to n do
begin
den := Y*So+Xo*exp(lln*tt[i]*(So+Xo/Y;
suml .- suml + sqr(So) * (Y*So+xo)/sqr(den) ;
sum2 := sum2 + So*sqr(Y*So+Xo) * (So-11n*Xo*Xo*tt(i]/Y/Y*exp(11n*tt[i]* (So+
Xo/Y)/exp(3*ln(den;
sum3 .- sum3 + ss[i]*So/den;
sum4 := sum4 + ss[i]*(Y*So+Xo)*(Sa-lln*Xo*Xo*tt[iJ/Y/Y*exp(lln*tt[i]*(So+
Xo/Y)/sqr(den) ;

77
end;
fYy := suml - sum2 - sum3 + surn4;
end;
l
j procedure secY;
var Y_l,fY_l,fYn,Yl real;
begin

Y_l:=0.95*Yn;
fY_l:=fYY(Y_l) ;
repeat

fYn:=fYY(Yn);
Yl:=(Y l*fYn - Yn*fY_l)/(fYn - fY_l}:
y l:=Yn;
fY_l:=fYn:
Yn:=Yl:

srs := ssrs(Yn}:
writeln(Yn:8:6,' , ,11n:8:7,' , ,srs:10:6):

until (abs(fYY(Yl < 0.00000001):


end;

function fkdkd(A:real):real:
var gar : real;
begin
suml:=O: sum2:=0: surn3:=O; surn4:=O:
for i := 1 to m do
begin
sum1 :=surn1 + B/A*sqr(tx[i]}:
sum2 :=sum2 + C/A*tx[i]*lnA+Xo*exp(al*tx[i]})/(A+Xo:
sum3 :=sum3 ~ tx[i]*ln(Xo/xx[i]):
sum4 :=sum4 + sqr(tx[i]);
end;
gar := (surn1 - sum2 + su~3)/surn4:
if gar < 0 then fkdkd .- 0
else fkdkd := g~r;
end;

function fAA(A:real) :real:


begin
kd : = fkdkd (A) ;
sum1:=O; sum2:=0: surn3:=O: sum4:=0: sum5:=Oi sum6:=Oi sum7:=O; sum8:=Oi sum9:=
for i:=1 to m do
begin
sum1 := sum1 + B/A*(B/A-kd)*sqr(tx[i);
sum2 : = sIlm2 + C*B/A/A,"tx[ il *ln( (A+Xo*exp (al*tx[ i]}) / (A+Xo) ) ;
sum3 := sum3 + B/A*tx[i]*ln(Xo/xx[i]);
sum4 := sum4 + (B/A-kd)*tx[i]*C*Xo*(l-exp(al*tx[i]/(A+Xo)/(A+Xo*exp(al*tt[i
sum5 := sum5 + c*C*Xo*(l-exp(al*tx[i]/A/(A+Xo)/(A+Xo*exp(al*tX[l] *
ln( (A+"{o*exp(al*tx[i]) )/(A+Xo:
surn6 := sum6 + C*Xo*(l-exp(al*tx[i]/(A+Xo)/(A+Xo*exp(al*tX(l]*
ln (Xo/xx [ i J) ;
sum7 := sum7 + (B/A-kd)*tx[i]*C/A*lnA+Xo*exp(al*tx(i]/(A+Xo i
surnS := sum8 + sqr(C/A*lnA+Xo*exp(al*tx(i)/(A+Xo))):
sum9 := sum9 + C/A*lnA+Xo*exp(al*tx[i])/(A+Xo*ln(Xo/xx[i]) i
end;
fAA:= (suml - sum2 + sum3 + surn4 - sum5 + sum6 - sum7 + sum8 - sum9) i
end;

78
,
'1

procedure s~cA:
var A l,fA l,fAn,Al real;
begin- -

t A 1:=0.98*An;
fA- l:=fAA(A 1);
-
repeat
fAn:=fAA(An) ;
A1:=(A_l*fAn - An*fA_l}/(fAn - fA_l)i
A l:=An;
fA_l:=fAn;
An:=Al;
if An < yn*so then An := yn*so + 0.0001;
Ks:= (Yn*So+Xo)*So/(An-Yn*So);
mu := Iln*Ks;
srx := ssrx(kd,An);
WRITELN(An:8:5,' ',Ks:10:3,' ',mu:lO:4,' ',srx:lO:6};
until (abs (fAA (Al < O. 000000001} ;
kd := fkdkd(An);
end;
begin
captura;
J\osol :=0;
Yn := initial gess ;
lln := Lmin ;
llinc := Linc
lIn := lIn - llinc;
llfinal := Lmax;
An := initial gess;
repeat
lIn := lIn + llinc;
secY;
B := 11n*So*(Yn*So+Xo);
C := Yn*So;
al := Iln*(So+Xo/Yn);
secA;
Ks:= (Yn*So+Xo)*So/(An-Yn*So);
mu:=lln*Ks;
writeln(outfile,lln:18:1l,' ',Ks:10:4,' ',mu:lO:4,' ',Yn:10:4,' ',kd:10:4,' "
writeln(lln:8:6,' ',Ks: 10:4,' ',mu:10:4,' ',Yn:lO:4,' ',kd: 10:4,' ',srs: 12:9,'
close(outfile);
end.

79

J
1

Appendix C

N on- biodegradable substrate

Herein the computer program used to estimate the non-biodegradablc substrale is pre-

sented. The computer program was writtcn in Turbo PASCAl v. 4.0. [6J.

80
.J

proqram substrate;
var infile,outfile : text:
ss,tt : array[O 15] of real;
bn,s,t,b : real:
1 i,n : integer:
procedure captura;
begin
assign(infile,'c:\inputfile');
assign(outfile,'c:\outputfile'):
reset(infile);
n :- -1;
repeat
n := n + 1:
readln(infile,t,s):
ss[n] :- s;
tt[n] := t:
until eof(infile):
close(infile);
end:
function bb(b:real):reali
var suml,sum2 : real:
begin
suml := 0; sum2:=O:
for i := 1 to n do
begin
suml := suml + tt[i]*sqr(ss[0])*exp{-2*b*tt(i]):
sum2 := sum2 + tt(i]*ss(i]*ss[O)*exp(-b*tt(i]):
end:
bb := suml - sum2;
end;
function ssr(b:real):real:
var surol : real;
begin
surol := 0:
for i :=l. to n do
suml := suml + sqr(ss(i) - ss(O)*exp(-b*tt(i));
ssr := suml;
end;
procedure secant;
var b l,bl,fb l,fbn real;
begin- -
b l := 0.9*bn:
fb_l := bb Cb_l) ;
repeat
fbn := bb (bn) ;
bl := (b_l*fbn - bn*fb_l)/(fbn - fb_l):
b l := "n:
fb l := fbn;
bn-:= bl:
until absb1 - b_l)*10000) < 0.00001;
end;
procedure Mean:
var as,ssrr,nosolin,nosolinc,nosolmax real;
begin
rewrite (outfile) i

81
noso11n := Non-soluble initial;
noso11nc :=Non-soluble increment;
nosolmax := Non-soluble final:
bn ;= 0.05:
for i :: 0 to n do
ss[i] := ss[1] - noso11n;
repeat
for 1 := 0 ta n do
ss[1] := ss[i] - noso1inc;
noso11n := noso11n + nosolinc;
secant;
ssrr := ssr(bn):
wr1teln(noso11n:8:3,' ',bn:15:12,' ',ssrr:20:10);
wr1teln(outfile,nosol1n:8:3,' ',bn:15:12,' ',ssrr:20:10);
until nosolin >= nosolmax;
close (outfile) :
end;
begin
captura:
Mean;
end.

(
82
,
f

Appendix D

SSRS and ~f) vs Y

The computer program presented herein was used to estimate the initial guesscs for Y and

to verify the feasible roots of FI' The computer program was written in Turbo PASCAl

v. 4.0 [6].

83
program Fl;
var infile,outfile : text~
tt,ss,xx : array[O 20] of real;
srs,srx,suml,sum2,sum3,sum4,sum5,sum6,sum7,sum8,sum9,nos01 real;
al,Ks,m,Xo,lln,Yn,Y,Yinc,Yfinal,So,t,s,x : real:
i,n : integer;
procedure captura;
begin
assign(infile,'c:\input.txt');
assign(outfile,'c:\output.txt');
reset(infile);
n:=O;
readln(infile,n);
for i := 0 to n do 4

begin
readln(infile,t,s) ;
tt[i] :=t;
ss( i] : =s;
end;
close(infile);
rewrite(outfile);
end;
function ssrs(Y:real):real~
begin
suml := 0;
for i:= l to n do
suml := suml + sqr(y*so+Xo)*So)/(y*so+Xo*exp(lln*tt[i]*(So+Xo/Y)-ss[i});
ssrs := suml;
end:
function fYY(Y:real):real;
var den: real;
begin
suml:=O; sum2:=0; sum3:=Oi sum4:=O;
for i := 1 to n do
cegin
den := Y*So+Xo*exp{lln*tt(i]*(So+Xo/Y~
suml := suml + sqr(So)*(Y*So+Xo)/sqr(den);
sum2 := sum2 + so*sqr(y*so+xo)*(So-11n*Xo*Xo*tt(i)/Y/Y*exp(11n*tt(i]* (So+
Xo/Y)/exp(3*ln(den ;
sum3 := sum3 + ss(i]*So/den;
sum4 := sum4 + sS(i]*(Y*So+Xo)*(So-lln*Xo*Xo*tt(i]/Y/Y*exp(lln*tt[i]*(So+
Xo/Y)/sqr(den);
end;
fyy := suml - sum2 - sum3 + sum4;
end;
begin
captura;
lIn := given value of L;
Yn := Initial value of Yi
Yinc := Increment of y;
Yn := Yn - Yinc;
Yfinal := Final value of y;
repeat
Yn := Yn + Yinc;
writeIn(outfile,Yn:l8:ll,' ',fYY(Yn):2l:11,' ',ssrs(Yn):12:9);
writeIn(Yn:18:11,' ',fYY(Yn):21:11,' ',ssrs(Yn):12:9);
until Yn >= Yfinal;
f close(outfile);
end.
84
Appendix E

SSRX and ~FJ vs A

The computer program presented herp!n was uscd to estimate the initial gucsses for A

and to verify the feasible roots of F2 - The program Wa'5 written in Turbo PASCAl v. -1.0

[6] _

85
program F2:{definitive}
var infile,outfile : text:
tx,tt,ss,xx : array(0 . 20] of real;
srs,srx,sumJ,sum2,sum3,sum4,sum5,sum6,sum7,~um8,sum9,nos01 : real;
1 A,B,C,Xo,An,llinc,llfinal,lln,al,Yn,Y,So,Ks,mu,kd,t,s,x : real;
i,n,m : integer;
procedure captura:
begin
assign(infile,'c:\input.txt');
assign(outfile,'c:\output.txt') ;
reset (infiIe) ;
n:=O:
m:=O;
readln(infile,n,m);
for i := 0 to n do
begin
readln(infile,t,s);
tt (il: =t;
ss [il :=s;
end:
for i := 0 to m do
begin
readln(infile,t,x);
. tx [il :=t;
xx [i] :=x;
end;
close(infile);
rewrite(outfile) :
end;
function ssrx(kd,A:real):real;
begin
suml : = 0;
for i:=1 to n do
suml := suml + sqr(ln(Xo*expBjA-kd)*tx(i]-CjA*lnA+Xo*exp(al*tx(i)/(A+Xo
ssrx := suml;
end;
function fkdkd(A:real):real;
var gar : real;
begin
sum1:=0: sum2:=0: sum3:=O: sum4:=0:
for i := 1 to m do
begin
suml :=s'lml + BjA*sqr(tx[i));
sum2 :=sum2 + CjA*tx(i]*lnA+Xo*exp(al*tx[i]/(A+Xo:
sum3 :=5um3 + tx[i]*ln(Xojxx[i]):
sum4 :=sum4 + sqr(tx(i) ;
end:
gar := (sumi - sum2 + sum3)jsum4;
if gar < 0 then fkdkd := 0
eise fkdkd := gar;
end;
function fAA(A:real) :real;
begin
kd : = fkdkd (A) ;
suml:=O: sum2:=O: sum3:=0; sum4:=0; sum5:=O; sum6:=O: sum7:=0: sum8:=O; sum9:=
r for i:=1 to m do
...

86
begin
suml := sum1 + B/A*CB/A-kd)*sqr(tx[i]):
sum2 .-sum2 + C*B/A/A*tx[i]*lnA+Xo*exp(al*tx[i)/(A+Xo));
sum3 := sum3 + B/A*tx[i)*lnCXo/xx[i);
), .-
sum4 .- sum4 + (B/A-kd)*tx[ij*C*Xo*(l-exp(alwtx(i]/(A+Xo}/(A+Xo*exp(al*tt[i
sum5 := sum5 + C*C*Xo*C1-exp(al*tx(i]/A/(A+Xo)/(A+Xo*exp(al*tx[i])} *
lnA+Xo*expCal*tx[i)/(A+Xo)) ;
sum6 := sum6 + C*Xo*(l-exp(al*tx(i]/(A+Xo)/(A+xo*eXp(al*tx[i))*
ln(Xo/xx[iJ) i
sum7 := sum7 + CB/A-kd)*tx[iJ*C/A*lnA+Xo*exp(al*tx(i)/(A+Xo));
sumS := sumS + sqr(C/A*lnA+Xo*exp(al*tx[i]j(A+Xo);
sum9 := sum9 + C/A*lnA+Xo*exp(al*tx[i)j(A+Xo))*ln(Xojxx[i]);
end:
fAA:= (sum1 - sum2 + sum3 + sum4 - sumS + sum6 - sum7 + sum8 - sum9) ;
end:
beqin
captura;
Yn := qiven value for Yi
lln := qiven value for L:
B := Iln*So*(Yn*So+Xo):
C := Yn*Soi
al := lln*(So+Xo/Yn);
An := Initial value for A;
Aninc := Increment;
An := An - Aninc;
Anfinal := Final value for A;
repeat
An := An + Aninc:
writeln(outfile,An:1S:11,' ',fAA(An):21:11,' ',ssrx(fkdkd(An),An):21:11)i
wri teln (An: 18: 11,' " fAA (An) : 21: 11,' ',ssrx (fkdkd (An) , An) : 21: 11) ;
," until An >= Anfinal;
close (outfile) :
end.

87
1

Appendix F

Analysis for the roots

In the [ollowing figures the SSR's and the roots that correspond to the minimum SSR's are

shown al the mlTllmUm znlersectioTl poznt for the Braha and Hafner (1987) [i] data sets,

and for the data sets from the experiments perforrned in this study. It can he ohserved

that in sorne graphs (Figs. F.I and F.3 to F.6) that a more or less sud den s!ope change

tales place ln the *" vs A plots. During the computations it was obs'r ved (on the screen

of tlw llIorlltor) that the slope change occurs when the non-negative constraint 011 kd

be(omcs active.

88
1

...

ili S
a il CO
III
...J nt:.
s:<
-
... f. ....J
...
/
-
..2
,
'"
;-

>-:
Ct::
ijJ
rr,
:::
~
... '" ... 0
N ... .. .,1
'"1

.,.... ..
"
on
'"
al

"0 ~
.." on
"' ......

-<:
'" ""c>
1 1 1 0
"" .,'"'"'"
0 0 <> 0 C>

(-JOI .. wu)
YI>/l:JP
0
0
<>
0
"
XIISS
" 0

"
C>

'"<> .'"
"'l" .
~~

.,...
.r

-
t..rJ
,...,
,or,
t f..

.....
...
if;

~~

N
::.:...
N
~
...- ,. ~
fi cf,
-1
ID ID
::...

ID
'"o
o
ID ...
<> ~
1
~
1
s:1 .
<>
1
e ...
"
1
0

...
<>
1
"
al

<>
7
"

(.puo'"O~J)
s~s

89
~

LB2
~".~

L82
-
100

50

j /
/
-1

-2

,;;: :n1
~

...
-100]

-150 1 ,-
~:::
I;! ..
v-
E
-3

-0

-5
-200 -i 1
-6
-250

-JOO

-J50
-1

1 -7

-9

-9

00 OS 12 16 -10
24 2B
-06 -04 -0 2 0 02 00 06 OS
Y 12 10
(ThauSOndS)
...
ao
000059

(0 000059
0 70 -1 / 000057
000056
60 -1 ./
000055
000054

",e:
~:
"'il
.
~

t
:j
JO / )(

!li
III
000053
000052
000051
00005
000049
000049
20 -1 \ /
000047
000006
ID -1 \ /
0000'5
000044
o 1 1'-- i ----= 000043
04 OB 12 16 2 20 2B
-06 -0 -02 0 02 0' 06 OS 12 1.
Y (ThauSOndS)
...

Figure F.2: lli


J)' vs y ,','
':;':;Re.,'
vs }' , f2f2.
,)" vs A an d ........
C,'C,'RX vs A Cor LB2
-~----- -- 2llI

. li ......
LB3
LB3
100 ]

~
000001
50
a
a -0 00001

-0 00002
-50
-000003

-100 -0 00004
"-
;;: ~-o 00005
... -150 ~
~-o 00006
-200 -000007

-0 oooos

-250 JI -0 00009
-lOO
-0 0001 j \
-350 1 -0000' 1

05 15 25 J5 -0 00012
y -0 6 -0 4 -0 2 a 02 04 06 08
(Thouao_>

'"
140 OOOOU

:::~ ::::::~
~ IJO

- / 1 \
100 000041
90
- 0000.
: 80
~c x
!if ; 70 ~ a 00039
~o ~

. 60 a 0003!
50
40 OOOOJ7

:: J\ 1
/ 000036

000035
'D
a 1 ? i~ 0~~4

05 15 25 35 -0 6 -0 -0 2 a 02 C4 C6 ce
(lhou>CndS)
y '"

Figurt' F :3: ':!.fi


(Il
vs} .~SRS VS } 1t VS A and ssn.\: \'5 A for LB3
~

LB4
JI.. '"
..
LB4
000001

'::j ~ -0 00001

-50 / -0 00002

-0 OOOOl ~
>- -0 00004 /
/
-100]
"
"- ~
~ ;:;-000005
" -150 ....

j " -OOOOO!
-200

-250 1 -000007

-000008

-lOO J1 -000009

-0 0001
-350
04 -000011
08 12 16 2 24 28 J2 -06 -04 -02 a 02 04 06 08
y (Thau"'_l
A

100
(,0 a 000~8
t-..:)
90
0000"
80
000056
70

.
000055
~
60
000054
"
Ille:
21::
IIIg 50
g 0.00053
t 40 '"
0.00052

JO
000051
20
00005
la
000049

a
000048
04 08 12 16 2 24 28 J2 -0 6 -04 -0 2 a 02 04 06 08
y (Thauoa_l
A

Figure F.4: ~ vs l", S'SRS vs Y, 9ft vs A and SSRX vs A for LB4


<;; ~ ". ~
...
LB5 LB5
l

=i
20
~ 2

0
-20 0

-40 -1
-60 -2
~
...
)-
-BO ~.!.
"-
;;:
... -100
-\20
~
V"
.
"-~

E
-3

-4

-\40 -5

-150 -5

-IBO -7
-200 -8
-220 -9
-240
-\0
0.4 08 \2 16 2 24 28 32
-06 -0 4 -0 2 a 02 04 06 03
y (Tl\......fIds)
Jo.

::j
0.0007
(.0
~ 000068

50 / 000066

000064

000062
~

:] \ /
en" 0.0006
~::
III~ ~
III 0 000~8

o OOO~6
20
000054

1 000052

I~ L
la
00005
o !
04 08 12 16
000048
24 28 32
-06 -0 4 -02 0 02 0.4 06 08
y
..
(Tl\OUlGndI)

Fi~lIre F.5: ~ VS }", SS'RS' VS Y, ~I~ VS A and S'SRX vs A for LB5


... ". . "\ .fi..-:A

L86 L86
OoO.J------T-----------------------____________________,
"1 _1 00012
'0 a DO"
0001
5 00009
o oooa
00001
00006
~ 0005

-5 00004
~ ;S a 0003
~
-=v -la
!;;
00002
00001
Q
==-- ,
-l' -0 0001
-Q 00Q2
-Q 0003
-20
-0 0004
-00005
-25 -00006
-00M1

-Jo;I----~--~----._--_r----._--_.--~r_--,_--~
3
-O.OOOIS
-0.0009 -1 I-----l----.------.----r----r----.--__r-_--.___-I
5 9 \1 -\Il la JO 50 10
(Th ...... ndI)
'f
A

OOllr--~r-------------------------- ------------,
~
,;:..
=j
35
0009

D.DDI

JO 0.007
';'
v 25 0006
Ole
2f~
.,"
6
o
20 i 0.005

0004
.5

0.003
10

0.002
5
0001
01 1~i
3 5 9 0;1----;---~----~---r----.---_r--~r_--~--_4
y
" -;10 la JO 50 10
(Th......ndll
A

Figure F.6: ~ vs Y, SSRS vs Y, Wf vs A and SSRX vs A for LB6


......................~------------~----------~~~~~'~-~ ,~
'~ .........r, -~ ...-~~ ~~ .. ~ ..... - - . : '....... Tf" ... '~- "'i"'~yr:;r -.~t,,,"~,,,
- "'..>r;I""'" . . ~.....~.-

! .....
LB7
80
LB7
&0 04
02
40
0
2Q
-0 2
a
-0 ,
-20
-06
-40
~ -08
>-
-50 1
~
... -80
~ ...
,~
-1
v -12
-100 ~"
v"
E -1 ,
-120 e -1 &
-140 -18
-150: -2
-180 -22
-200 -2 ,

-220 -25
04 08 12 16 2 24 28 -28
-&00 -400 -200 Q lOG 40G
Y 600 800
A

4~

0.00059

40
/ 1
c.o a 0005S
0- J~ J /'

30
/ 000057
";
v 25
<ne:
5~
~
X
fil ~
&;

C
20 / !li
'"
00005&

15
000055

"]
5

0
\ / 000054

04 OS 12 000053
16 24 28
-600 -400 -200 a 200 400 &00 SOO
y
A

Figlllt:' F.i: ~ VS L .-;SRS VS y', ~~tt VS A and SSRX vs A for LB7


"'~ ....,.i:<\
..
Run 1
Run 1
:J ---
COD
01
0
COD

------
-01
lO -i /
-02
20 -0 J
-04
10 -0 ,
~
,"
~
0
~.!.
,2
-0.6
-07
-=" -10 ~. -0 Il
v -09
-20
-1
-lO -II
-1 2

~:11 -1 3
-1.4
-1.5
-60 1
-10
0.1 03 0' 07 09 Il 13 15 17 19 21 -06 -04 -02 02 04 06 011
y (ThOU .._ )
A

10 DDDD61,--------------------,------______________________________- ,
c:.o
(j)
o 00D609
9
00006011
0000607

l
Il
o OD0606
7 o DDD6D5

~.~~~:~~
';' 6
v
Ille:
5:: x
5
III~ ~O 000601
00006
4
0000599 /
0.000598
J
0000597
2 o 00D596
0000595
0000594
0.00059l
0+1~-r.-,-~._r-. .~_._r_.,_~._._. .~ 00005924---r-~--,__,r_~--;__,r__r--~_,--_r--~_,--~--~~--~
0.1 03 05 07 09 Il 13 15 17 19 21 -06 -04 -02 02 04 06 011
(Thou .. _ )
y
A

Figure F.S: *'" vs Y, SSRS vs 'V, ~ vs A and SSRX vs A for Run 1 (COD)
i .~ ~ ..

Run 1 Run 1
TOC
4 TOC
01
l
a
2
-0.1

0 -0.2
-1
-0 l
-2
-3
"<II -0 ~

~ o!.
~:
-0.5
"-
;:: -~
v -06
-5
"
-5 -07
-7
-0 B
-1
-09
-9
-la -1

-11 -1.1
-12
-12
Q~ 15 25 J5 45 55 -IBD -ua -100 -60 -20 2Q 60 100 140 180
Y A

00005989
600
CO 00005988
-.J 00005987

/
00005986
500 -l
00005985
a 00059!l~
400 -l 00005983
00005982
i 0005981
~ /
VI
.,~ JOO '" 0000598
00005979
00005978
200 -1 / 00005977
00005976
00005975
100 -1\ / 00005974
00005973
00005972
o 1 -IBO -140 -100 -60 -20 2Q 60 100 ..0 IBO
05 15 25 35 45 55
y
...

Figure F.9: ~ vs },., S'SRS' vs }", 7ft vs A and SSRX vs A for Run 1 (TOC)
.. ~ ~~ ..
Run 2 Run 2
COD

~~
COD
1 1
:so
40
~ 0

:~j
-1

,.v
"-
.:v /
J
~.1
,~
~
v"
~

.
E
E
-2

-3

-4

-"V
-20

-30

-40
-~

-6

-50
-7
0.3 0' 07 09 Il 13 1~ 17 19 21 -1111 -140 -100 -60 -20 20 60 100 140 180
Y
A

I~
14
1 00020111 ------
~ 13..1 /' 0002017
00
0002016

/
12 ]
Il

10 0002015

-;-

:~
0002014
Ole
5f~
" x
.,~ gfo 002013
t. (Il

0002012

0002011

J
0.00201

/'
1 1 1
0002009

03 05 07 09 1.1
-1 0002008
1J 15 17 19 21 -160 -140 -100 -60 -20 20 60 100 140 1110
Y A

Figure F .10: Wvs Y, SSIlS vs }", i?ft vs A and SSRX vs A for Run 2 (COD)
;. " .....
Run 2
, i
TOC Run 2
TOC
01

-01
Ji / -02

-03

'1 /
-0 ~
>- ~
"
"-
;;: ~ ...
1 -05
... ,~
f! ....E
-0.6

-07
E
-08
-09
o 1
-1

-II

-12
-1
0 ~ 6 -1 J
B ID
-1110 -140 -100 -60 -20 20 60 100 140 180
Y
A

19
o OOJ9J~
<.0 19
c.o 17 0003933

16 o 00J932
15 0003931
14
000393
13
0003929
~

.; c
.
"
12
Il
0003928
0003927
~~ X
"'5s.: 09 ero 003926
Cil
t 011 o 003925
07 0003924
06
o 00392~
05
04 0003922
03 0003921
02 o 00J92
0\ 0003919
o 00Ja'8
0 2 10 -1110 -140 -100 -60 -20 20 60 100 140 IBO
Y
A

Fi gUI (' F.l/. ~ VS Y, SSHS \'5 }', 4f VS A and SSRX vs A for Run 2 (TOC)
"'",. " il>- '"
...
Run 3 Run 3
COD
60 COD
a oooos
50
40
~ 0
JO
20, oooos
./ -0
la
0 -00001
/
-la

_n~
-0 00015

/
~
'!7
"- -'0 ~
;:: ~ -0 0002
v
-40
....v
-'0
-60 -00002'
-7Q
-80 J 1 -0 0001
-90
-100 -' 1 -0000J5
-lia
-0 0004
-120
-130
OJ 05 07
.
09
-r-I

Il 1J 15 17 19 21
-0 00045
-IBO -140 -100 -50 -20 2D 50 100 140 IBO
Y
A-

0001
~

-
15
14
./ 000295
0 13
0 00029
12 ]
1\ / 000'85

.....
10
Q 002!
9
utC:
)( 000275
8
~:
ut~ ~ 00021
s=
t 6
000265
5
00026
4

l 000255
2
00025

o , 0002"
-IBO -140 -100 -50 -20 2D 50 100 140 IBO
01 OS 07 09 Il 1J 15 17 19 2.
Y A

Figure F.12: ~ vs Y, SSRS vs F, fJff vs A and SSRX vs A for Run 3 (COD)


____________________________ ... ..... .. ..... .. .. ~... 'I.... .. ~I~~~.,... --~ .. i:~'l''''3l'--'#l1 ..,(J1't"'

~
_
=
.
.
.
,
.
.
-
"
.
.
.
.
.
.
-
~
.
,
.
-
.
,
.
.
.
,
~
~
~
"
"
~
.
:
r
,
.

~

y - .... .,.. . . "<',.,""''l:IIr''~..,.. .... p.....'t .. -.:~''t 11'.,,"J<~ ... '";/"'f""~~
'i ....t ..

" <ri ....


Run 3
TOC Run 3
'.5 ~OC
0000.

05
- -0 000\
0

-05
-0.0002
>-
v
"-
~
-. ~
~ -00003
"-
v
-. :1

-0 000'
-2

-25 -00005

-3

-.J
-0 0006

-35
2 '0 '4 22 26 JO -00007
'8
y -'BO -'40 -'00 -60 -20 20 60 '00 140 'SO

0.0073
...... 800 ]
0
......
700
/ 1 00072

00071

... ~
/
0007
500

)(
00069
III
5
III
400 il!
III
00068

300 -i /
00067

:::]J 10 '8 Z2 26 JO
00066

00065

00064
-.80 -'40 -'00 -60 -20 20 60 '00 '40 '80
" y
'"
Flg\lle F.13: *" VS }'. SSRS vs }'. ~ vs A and SSRX vs A for Run 3 (TOC)
.. ~ ..
Run 4
05
COD Run 4
05
04

03

02
o1 -05

.
0
-1
-0.1
'" ~
t~
,:;:
~~
v
-02
-03

,~
-1.5

~.
-04 v
E -2
-05
-25

-"u
-06

-3
-OB
-09 -35
-1
0 02 04 06
-02 0 02
Y
04 06 OS
(ThOU..,,.,')
...
o 00017

~ Ai 000036

/
0 50
~

000035
40

~i
t
...
~c:

lO i / ~
CIl
a 00014

o 00013

20
o 00012

o OOOll

':1 0 02 04 06
00003
-0 2 0 02 04
(Thou ...,.,.)
06 OB
y ...

Figure F.14: ~ vs V. SSRS vs Y, ~ vs A and SSRX vs A for Run <1 (COD)

t l; '!I ...
Run 4 Run 4
TOC
12 TOC
0.5
10

:~
0

2
/ -05

-1
0
~
~ -2 ~,!, -1.5
~
...,. ,2
-4 ~.
" -2
-&

-8
-25
-10
-12 -3
-14
-16 J -35

-18
-4
04 011 12 1& 2 24 28 -110 -90 -70 -50 -30 -10 10 JO ~ 70 90 110
Y

"
13 0000373

....- 12 -! / 00001725
0
~
I l -, ;-
0000372

00003715

.
~

)( 0 000371
tI>" "
~

"1
~:: 08
tI>~ II>
00003705
"
t- 07
000037
06
o 0003695
05

0000369
04

03 00003685
O' OS 12 16 2 28 -110 -90 -70 -50 -30 -'fJ ICI JO ~ 70 90 lI0
y ..

Fig\llc F.15: *" VS } ' , SSRS VS }'", 9ft VS A and SSRX vs A for Run 4 (TOC)

~ ... "iIL- _ _ . " ' _


~
.,,,~

.-'~<\

Run 5 Run 5
COD COD
60 0001
~O

3D
0
r---
20
10 -0 001
0
,.v
~
-10
) -0002
"-
v -20
v i
-JO
-~ -0003
-~O ~

-60
-0 004

.
1

-70

-Ba 1

-90 -0 oo~ -

02 04 06 oa 12 14 16 1 a 2 -1 3 $ 7 11 13 1$
y
(Th ...... _>
A

III
00011
...... \7
0 16
~ 0007
I~

\4
13 0006
\2

';'
Il o OO~
fI'''" 10
~~
.... il
~
t
9
Il
7
..~ 0004

a OOl J

1~
4 Q 002
3
2
1 000:
o-i-r-
04 06 OB 12 14 16 lB 2 -1 1 3 $ 7 9 11 13
Y (Thau ... _> 15

Figure F.16: * vs Y, S'SliS vs Y, 2ft vs A a.nd SSRX vs A for Run 5 (COD)


">
-
Run 5 Run 5
TOC 01
roc
1
6 0

5 -01
4 -02
3
-0.3
2 ~
1 -04
~"'
,2
~
..... 0 /:!.
v
-05

-1 -0.6
E
-2 -0.7
-3
-0 8
-4

-5
-6
-0 9

-1
.
-II
-7
0.4 OB 12 15
-200 -100 o 100 200
24 2B 32
'(
...

00004'5

-
CI
CJt
"'j
800 / 1
0000454

0000453

0000452

0000'"
000045
700
1 / 00009
., )(

.,fl 600 i / eio


., 0008
0000447

0000446
500~ /
00005

/' o 0~O4&4
400-l /' 0000443

-~ o 000U2
a OOOUI
JOO 200
-200 -100 100
o. 08 1l 16 2 2B J 2
...

Flgtlrt~ FA: ~ vs }', SSH8 vs Y, W1 vs A and SSRX vs A for Run .5 (TOC)


Bibliography

[1] American Public Health Association, American Water Works Association. and Wa-

ter Pollution Control Federation. Standard Methods for the Examination of Water

and Wastewater. American Public Health Association, American Water Works As-

soeiation, and Water Pollution Control Federation., Washington D.C., 14 edition,

1976.

[2] G.G. Bach. Numerical A nalysis. McGill Urlversity, Montreal, 1982.

1:3] L.D. Benefield and C.W Randall. Biological Process Design for Wastewater Treat-
ment. Prentiee-Hall Ine., Englewood Cliffs, N.J., 1980.

14] J. Blok. l\leasurements of the viable biomass concentration in activated sludge by


respirometric techniques. Water Research, 10:919-925, 1976.

[5] J. Blok. Respirometric measurements on activated sludge. Water Research, 8:11-18,

t9l,

[6] Ine. Borland. TU7'bo Pascal v. 4.0. Borland International., Scotts Valley, Ca, 1987.

[il A. Braha and F. Hafner. Use of lab batch reactors to model biokinetics. Water
RC8carch, 21(1):3-81, 1987.

[8] Il. B. Braun and M. Berthouex. Analysis of Lag Phase BOD Curves using the Monod

Equations. Waler ResouTees Research, 6:838-844, 1970.

106
1
[9] A.C. Camara and Randall C.W. Theoretical estimation of kinetic pararncters in the

design of activated sIudge processes. Journal Water Pollution Control Federation,

56(4):388-389, 1984.

[10] W.E. Gates and J.T. Marlar. Graphical analysis of Batch Culture Data Using the

Monod Expressions. Journal Water Polluhon Control Federation, 40:469, 1968.

[11] A.F. Gaudy, Jr., P.Y. Yang, and A.W. Obayashy. Studies on the total oxidatioll cl

activated sludge with and without hydrolytic pretreatment. Journal Water Pollution

Control Federation, 43{l):40-54, 1971.

[12] A.F. Gaudy Jr., P.Y. Ramanathan, and T.V. DeCeare. Studies on the opcrational

stability of the extended aeration process. Journal Water Pollution Con/roi Fedcra-

tion, 42(2):165-179,1970.

(13] D. Herbert. Recent progress in microbiology. In G. Tunevall, Alquimist and Wikscll,


.'
editor, VU International Congress for Microbioloqy, page 381, Stockholm, 1958

[14] James M. Montgomery, Consulting Engineers, Inc. Water Treatment, Pnnczplcs ami

Design. John Wiley., N.Y., 1985.

[15] J.L. Keuster and J.H. Mize. Optimization Techniques wzth FORTRAN. McGraw-HiIl,

N.Y., 1973.

[16] G. Knowles, A.L. Dowing, and M.J. Barrett. Determination of kinctic constants for

nitrifying in mixed culture, with aid of an electronic computer. Gen. Mlcrob20I.,

38:263-278, 1965.

[17] A.'W. Lawrence and P.L. McCarty. A unified basis for biological trcatm{'nt desigll

and operation. Journal of the Sanitary Engi1lccrzng Dzvzszon, ASCE., 96(6):7.57-778,

1970.

107
,
[18] D.W. Marquardt. An algorithm for least-squares estimation of nonlinear parameters.

Soc. [ndust. Applied Mathematics, 11(6):431-441, 1963.

[19] Metcalf and Eddy Inc. Wastewater Engineering Treatment Disposai Reuse. McGraw-

Hdl, N. V., 2 edition, 1979.

[20J J. Monod. The growth of bacterial cultures. Annual Review of Microbiology, 3:371-
394, 1949.

[21] F. Moser, H. Schnitzer, and O. Wolfbauer. Theoretischer Vergleich der Leistung und

des 8etriebsverhaltens von Uings- und Mischbecken fr Abwasserreinigungsanlagen.

Das Gas Wasserfach, 114:326-330, 1973.

[22] A. W. Obrayashi and A.F. Gaudy, Jr. Aerobic digestion of extracellular microbi~\l

polysaccharides. Journal Water Pollution Control Federatwn, 45(7):1584-1594, 1973.

( [23] Joint Committee of the Water Pollution Control Federation and the ASCE. ~:astew

ater Treatment Plant Design. Manual of Practice No 8. Washington D.C., 1977.

[24J S.J. Pirt. Princzples of .Hzcrobe and Cell Cultivation. John Wiley and Sons, Inc., 1
N.Y., 1975.

[25J T.D. Reynolds and J.T. Yang. Model of the completely-mixed activated sludge pro-
cess. In 21st Purdue Ind. Waste Conf, pages 696-713, 1966.

[26J J.A. Robinson and W.G. Characklis. Simultaneous Estimation of Vma:r, [(m, and

the Rate of Endogenous Substrate Production (R) from Substrate Depletion Data.

Mlc/'Obial Ecology, 10:165-178, 1984.

[2] .1.A. Robinson and J.M. Tiedje. Nonlinear Estimation of Monod Growth Kinetic

Parameters from a Single Substrate Depletion Curve. Applied and Environmental


(
:\11 Cl'ob 1Ology, 45(5):1453-1458, 1983.

108
[28] S. Simkins and M. Alexander. Models for mineralization kinetics with the variables

1 of substrate concentration and population density. Applied and Environmental 1\[,-

crobiology, 47(6):1299-1306, 1984.

[29] G. Tchobanoglous and E.D. Schroeder. Water Quality: Ch(lracteristics, 1\1odc/ing,

Modification. Addison-Wesley, Ine., Massachusets, 1985.

[30] L.L. Templeton and C.P.L. Grady Jr. Effect of culture history on the detcrmination of

biodegradation kinetic.s by batch and fed-batch techniques. Journal Water Pol/ution

Control Federation, 60(5):651-658, 1988.

[31] L.F. Tischler and W.W. Eckenfelder. Linear substrate removal in the activated sludge

process. In Proceedings of the Fourth International Conference on Water Pollutzol/

Research., Prague., 1969.

... [32] V.C. Uloth and D.S. Manavic. Aerobic bio-treatrnent of a high-strength leachate .
1<

Journal of the Environmental Engineering Division., 103(8):647-661, 19i7.

[33) A. van Niekerk, D. Jenkins, and M.G. Richard. Application of batch kinetic data to

the sizing of continuous-fiow activated sludge reactors. lVater Science and Tec/mo/-

ogy., 19:505--516, 1987.

[34] P.R. Vasicek. Use of a kinetic study to optimize the activated sludge proccss. Journal

Water Pollution Control Federation, 54(8): 1176-1184, 1982.

[35] P.Y. Yang and A.F. Gaudy Jr. Control of biological solids concentration in extcnded

aeration. Journal Water Pollution Control Federation, 46(3):543-553, 19i4.

109

Вам также может понравиться