Professional Documents
Culture Documents
ABSTRACT
ALIYAH BAIDA WIWIYANTI. Performance and Biokinetic Modeling of
Modified Upflow Anaerobic Sludge Blanket Unit in Treating Palm Oil Mill
Effluent. Supervised by ALLEN KURNIAWAN and MARK LARRACAS SIBAG.
The high levels of contaminants in Palm Mill Oil Effluent need to be processed with
a high-rate anaerobic digestion reactor, such as the Upflow Anaerobic Sludge
Blanket (UASB). Manual performance measurement of the UASB unit can be
laborious and resource-intensive. Therefore, this research aims to analyze the
performance of the UASB unit by implementing variations in Hydraulic Retention
Time (HRT), comparing the effluent estimation using the Monod Model and the
Artificial Neural Network (ANN), and estimating the biokinetic parameters of the
UASB unit under unsteady-state conditions. The research used a laboratory-scale
reactor from November 2022 to June 2023. It was determined that an HRT of 2
days in the UASB unit proved to be the optimal operational HRT, achieving 68%
removal of Chemical Oxygen Demand and 93% removal of Total Suspended
Solids. Comparing the effluent estimation models, the ANN provided the most
suitable estimation with a MAPE value of 17%. Furthermore, the biokinetic
parameters of the UASB unit were estimated using the Monod Model and obtained
values as follows: 𝐾e =0.1593 d-1, 𝜇max =1.887 d-1, 𝐾s =3459 mgCODs/L, 𝑌=0.98
mgTSS/mgCODs, 𝐾h =0.58 d-1, and 𝑘=0.0004 d-1.
Keywords : ACSt, ANN, Monod, POME, UASB
ABSTRAK
undergraduate thesis
as the requirement to obtain Bachelor’s Degree in
Civil and Environmental Engineering Department
Approved by
Supervisor:
Dr. Eng. Allen Kurniawan, S.T., M.T.
NIP. 19820729 201012 1 005
Co-supervisor:
Mark L. Sibag, Ph.D.
Known by
Head of Department:
Dr. Ir. Erizal, M.Agr. IPM
NIP. 19650106 199002 1 001
PREFACE
The author prays the praise and gratitude to the Almighty, Allah swt., who
has given the author health and blessing to properly complete the undergraduate
thesis entitled "Performance and Biokinetic Modeling of Modified Upflow
Anaerobic Sludge Blanket Unit in Treating Palm Oil Mill Effluent" right on time.
The thesis was created and submitted to fullfil the requirement for a bachelor’s
degree in the Department of Civil and Environmental Engineering, IPB University.
On this occasion, the author would like to thank all associates who gave their
support in the undergraduate thesis preparation, particularly to:
1. Allah SWT for all the blessing, grace, and mercy.
2. Aida Abbas and Siti Fatimah Dina as parents who have given unconditional and
unlimited love, support, and prayers.
3. Dr. Eng. Allen Kurniawan, S.T., M.T. and Mark L. Sibag, Ph.D., as supervisors
who have given suggestions and guidance from the beginning until the
completion of this thesis.
4. Maryam Abbas, Nur Fadillah Rahman, Siti Aisyah Abbas, Eka Audia Tiasa,
Muhammad Hakam Asy Syifa, Muhammad Habibi Ar Rayhan, and the entire
family for all the prayers, support, and encouragement.
5. Kadek Yulia Prameswasti, Dzaki Nauval, Kenannita, Oktavian Wahyu Pratama
Ajie, and Muhammad Faiz Kahendran as the ACSt teammate for all the
company, assistance, and advice.
6. Allifiya Salsabil Nugrohoputri, Dzaki Nauval, Rais Rahmadi, and Ahmad Rijani
Hasby whom have provided home during the most exhausting time of the
undergraduate period.
7. Owen Jacob Notonugroho, Fatihaturrizky Amelia, Mayandra Salsabhila Adam,
Humaira Amirani, Fauzan Fadhlurrahman, and friends from Allen's Supervised
team for all the helps in the completion of this thesis.
8. Chintia Dwiyundani Suharto, Nahda Kamila Assyifa, Maharani Bilqist Caroline,
Imma Nur Izzati Adzkia, Zayyaan Nabiila Khairunnisa, Gaizca Betha Bianca,
Aisyah, Alvian Setyo Nugroho, Hardiyan Yulianto, and all friends from
Navillera Atiharsa for the contributions during the lectures and the thesis
completion.
9. Lastly, all other friends and associates who can not be mentioned individually
for the help and support.
The author has contrived the thesis as best as possible, but the author is also
aware of many imperfections in both content and grammar. Consequently, the
author welcomes any suggestions and constructive corrections to improve. The
author wishes this thesis would be helpful in enriching readers’ knowledge.
CONTENTS
LIST OF TABLES x
LIST OF FIGURES x
LIST OF APPENDIX x
LIST OF NOTATIONS xi
I INTRODUCTION 1
1.1 Background 1
1.2 Problems Formulation 3
1.3 Main Objectives of The Research 3
1.4 Significance of The Research 3
1.5 Scope of The Research 3
II METHODOLOGY 4
2.1 Research Site and Timeline 4
2.2 Research Operation Procedure 4
III RESULT AND DISCUSSION 14
3.1 POME Characteristic Analysis and The UASB Unit Design 14
3.2 Biomass Seeding and Acclimatization 16
3.3 Performance Evaluation of the ACSt Reactor 17
3.4 The UASB Unit Performance Evaluation 20
3.5 The Effluent Concentration Estimation Modeling 23
3.6 Biokinetic Coefficient Parameter on The Monod Model 24
CONCLUSION 29
REFERENCES 30
APPENDIX 37
BIOGRAPHY 57
x
LIST OF TABLES
LIST OF FIGURES
LIST OF APPENDIX
LIST OF NOTATIONS
1 I INTRODUCTION
1.1 Background
Biological process units are the relatively low operational cost technology
used at most wastewater treatment facilities, compared to chemical and physical
treatment units. The biological process is primarily used to remove nutrients in
wastewater, such as dissolved organic matter, nitrogen, and phosphorus, using a
variety of microorganisms (Santos et al. 2020; Asadi and McPhedran 2021). The
aerobic biological process has been used over 60 years to remove organics from
wastewater (Nam et al. 2004). Hence, almost half of the electricity cost in
wastewater treatment facilities is caused by the aeration process in an aerobic
process (Liao et al. 2022). Moreover, the aerobic process is just commonly used to
treat less than 1.000 mg CODs/L wastewater (Amin Goli et al. 2019).
On the other hand, anaerobic process technology has been widely proven to
give valuable advantages rather to conventional aerobic treatment (Perendeci et al.
2012). High contaminants removal efficiency (for more than 4.000 mg CODs/L) in
low or high loading rates, low energy consumption, low requirement of nutrients
and chemicals, energy production (methane gas), and low clogging occurrences
become advantages of the anaerobic process (Anijiofor et al. 2017). However,
combined anaerobic-aerobic treatment gives many advantages, such as improving
sludge dewatering properties, providing additional solids reduction, reducing
nitrogen, and effectively removing organic matter from high-strength wastewater
(Zhou et al. 2006; Novak et al. 2011). The benefits of using a combined treatment
approach encouraged the adoption of an anaerobic process as the primary stage in
the treatment of highly organic wastewater. The effluent will be treated under
aerobic conditions to guarantee compliance with the required wastewater standards.
This strategy is recommended because of its increased efficiency and cost-
effectiveness in controlling concentrated wastewater, particularly in settings such
as rubber industry wastewater, pulp and paper industry wastewater, and Palm Oil
Mill Effluent (POME) (Mamińska 2017; Chung et al. 2018; Nasir et al. 2018).
Table 1.1 POME characteristics and Indonesian standard comparison
Parameter Unit Characteristic Standard
pH - 4-5 6-9
Biological Oxygen Demand (BOD) mg/L 25000-65714 100
Soluble Chemical Oxygen Demand (CODs) mg/L 44300-102696 350
Total Suspended Solids (TSS) mg/L 18000-46011 250
Oil and Grease (OG) mg/L 4000-9341 25
Total Nitrogen (TN) mg/L 750-770 50
The POME itself has been one of the significant wastewater problems in
Indonesia. The total of 15.08 M hectares of palm oil plantations in 2021 makes
Indonesia the world's biggest palm oil producer (Rizaty 2022). Estimated 5 to 7.5
tons of water will end up as POME for each ton of crude palm oil produced
(Fairuzah and Asnawi 2012). POME is characterized as a brownish, acidic, hot, and
viscous liquid with a high value of contaminant. The significant differences
between the raw POME characteristics and the effluent standard of POME from
Indonesian Ministry of Environment Regulation No. 5/2014 are nearly 300 times
different for the CODs parameter, as shown in Table 1.1 (KemenLH 2014; Zulkifli
2
2 II METHODOLOGY
discharge of the POME. The UASB unit was designed using the OLR of 60000
mg/l/day, the HRT of 48 hours, and the discharge of 40 L/day for the dimensional
design calculations. As a result, the UASB’s volume of 80 L was obtained.
The biomass mixed cultures microorganism for the UASB unit obtained from
Wastewater Treatment Plant II (WWTP II) Jababeka in the Oxidation Ditch (OD)
unit mixed with microbial starter. The microbial starter contained probiotic
anaerobic and aerobic microorganisms culture that stimulates the growth of
beneficial microorganisms in the form of suspended biofilm and degrades organic
compounds (Rahadi et al. 2018; Suryawan et al. 2023). The combination of existing
sludge and the microbial starter was chosen due to the previous study from
Kullavanijaya et al. (2022), who stated that the result of a rapid start-up, well
microbial activity, and high removal efficiency from co-inoculation of the Waste
Activated Sludge (WAS) mixed with an alternative source of a surplus supply of
mixed microbial cultures.
The seeding and acclimatization processes were necessary to increase and
prepare the biomass to become accustomed to utilize POME as the primary
substrate. These processes were conducted in a closed tank outside the reactor with
a capacity of 80 L. Monitoring of the TSS and CODs concentration was carried out
during the processes to measure the increase of biomass and the decrease of organic
compound pollutant in the form of CODs value (Budiastuti et al. 2023). The TSS
and CODs parameter sample was taken from the homogenous biomass inside the
tank after manual stirring. During seeding, glucose (C6H12O6) was given to the
biomass as the initial substrate using the food per microorganism (F/M) ratio of 0.2
(Jaouad et al. 2020). The assumption of substrate conversion between glucose and
6
CODs is 1 gram of glucose equal to 1.067 grams of CODs (Bjerre et al. 1996).
While for the acclimatization process, glucose was gradually replaced by POME
using the ratio of 0.25 POME; 0.5 POME; 0.75 POME; and completely using
POME at the end of the acclimatization process as the primary substrate source.
Hence, these processes also needed additional micronutrients such as nitrogen and
phosphate in the form of CH4N2O and P2O5 for biomass with the scale of 700:5:1
compared to the CODs value (Gutiérrez et al. 1999). After completing these
processes, the biomass was introduced into the reactor for the primary POME
treatment process in the ACSt reactor.
Start
Literature studies
Modeling
inaccurate
Statistic analysis
Modeling accurate
Finish
The ACSt reactor was fully operated to treat POME on April-July 2023 at
ambient room temperature. The reactor treatment process used ten days of
operational time for each Hydraulic Retention Time (HRT) variation under
unsteady state conditions. HRT variations in ACSt can be seen in Table 2.2. The
UASB unit uses HRT of 8, 4, and 2 days. The total of 7 sampling points in ACSt
reactor presented in Figure 2.3 with the sampling points of the influent of ACSt
reactor (and UASB unit), the effluent of UASB unit, the effluent of contact unit, the
effluent of ACSt Recator (and Sedimentation unit), the sludge underflow of
sedimentation unit, the effluent recirculating flow from stabilization unit, and the
settled biomass of UASB respectively. The measurements of the ACSt reactor used
the parameters of pH, Temperature, TDS, TSS, CODs, Ammonia, OG, and TN
concentrations. Table 2.3 details the measurement instruments and chemicals for
each parameter on ACSt measurements. Each parameter referred to SNI
6989.2:2009 for CODs Measurements with Closed Reflux Spectrophotometrically,
SNI 06-6989.30-2005 in Phenate Spectrophotometric Measurements for Ammonia,
SNI 06-6989.26-2005 in Gravimetric Measurement Method for Total Solids
Content, SNI 06-6989.10-2004 in Gravimetric Measurement Method for Oil and
Grease, and SNI 06-6989.52:2005 for Organic Nitrogen Measurements with Macro
Kjeldahl and Titration Method. The measurement result was then compared to the
POME effluent standard in Indonesia from Indonesian Ministry of Environment
Regulation No. 5/2014 to obtain the ACSt and UASB performance evaluation.
Table 2.2 Variations of Hydraulic Retention Time (HRT) in ACSt reactor
Variation UASB (d) IFAS (d) Stabilization (d) Clarifier (d)
1 8 0.67 0.25 0.21
2 8 0.67 0.21 0.21
3 8 0.67 0.17 0.21
4 8 0.50 0.17 0.21
5 8 0.33 0.17 0.21
6 4 0.33 0.17 0.21
7 2 0.33 0.17 0.21
Table 2.3 The instruments and chemicals for conducting of the research
Parameters Instruments Chemicals
pH pH meter Solution pH 4.01 and 6.86
Temp. TDS and Temperature Meter
TDS TDS and Temperature Meter
TSS - Analytical Balance - Whatman Ashless Filter Paper
- Porcelain Cup 42 Diameter 55 mm
- Desiccator - Distilled Water
- Oven
- Filter Paper Clamp
- Filter Tool
- Suction Pump
CODs - Spectrophotometer - Distilled water
- Cuvette - K2Cr2O7
- Digestion Vessels - H2SO4
- Block Heater - Ag2SO4
- Burette - NH2SO3H
- Volumetric Flask - HOOCC6H4COOK
- Volumetric Pipette
- Beaker Glass
- Magnetic Stirrer
- Analytical Balance
Ammonia - Spectrophotometer - Distilled Water
- Cuvette - NH4Cl
- Analytical Balance - C6H5OH
- Erlenmeyer - C5FeN6Na2O
- Volumetric Flask - C6H5Na3O7
- Volumetric Pipette - NaClO
- Measuring Pipettes
- Beaker Glass
OG - Analytical Balance - Distilled Water
- Separatory Funnel - HCl or H2SO4
- Distillation Flask - Organic Solvents
- Glass Funnel - n-Heaxane
- Filter Paper - Methyl Tert Buthyl Ether
- Centrifugal Device (MTBE)
- Vacuum Pump - Na2SO4 Anhidrat
- Distillation Adapter - Dichloro Methane (DMC)
- Water Bath
- Desiccator
TN - Kjeldahl Apparatus - Distilled Water
- Heater - K2SO4
- Distillation Equipment - CuSO4
- Glassware - NaOH
- pH Meter - Na2B4O7
- Analytical Balance - H3BO3
- Ovens - Methyl Orange
- H2SO4
- Na2CO3
Where 𝜇 is the specific growth rate (d-1), 𝜇max is the maximum specific growth rate
or the increasing concentrations rate limit (d−1), 𝐾s is the Monod saturation constant
or the substrate effluent concentration at half of the maximum rate (g CODs L−1),
and 𝑆an is the concentration of substrate effluent (g CODs l-1).
Furthermore, the Monod Equation needs to be combined and modified with
other equations to predict the effluent of the UASB unit. Firstly, the rate of change
in biomass from Eq. (2) was firstly modified into Eq. (3) to find the specific growth
rate (𝜇) from the biomass balance approach. While the sludge production equation
from Eq. (4) was modified into Eq. (5) to estimate biomass concentration inside the
reactor (𝑋an ).
𝑑𝑋an 𝑉an = 𝑉an 𝑋an 𝜇 𝑑𝑡 − 𝑉an 𝐾e 𝑋an 𝑑𝑡 − 𝑄an 𝑋an eff 𝑑𝑡 (2)
𝑋an (t1)
ln
𝑋an (t0) 𝑋an eff
𝜇= (𝑡1 −𝑡0 )
+ 𝐾e + 𝑋 (3)
an (t1) 𝜃i an
𝑑𝑋 𝑑𝑆
= 𝑌 𝑑𝑡 (4)
𝑑𝑡
𝑋an = 𝑌 (𝑆in − 𝑆an ) + 𝑋an eff (5)
-1 −1
Where 𝐾e is the constant of death rate (d ), 𝑌 is yield coefficient (g TSS g CODs ),
𝑡 is the day of the experiment (d), 𝑆in is the concentration of substrate influent (g
CODs l-1), 𝑋an & 𝑋an eff are biomass concentrations in the UASB and the effluent
of UASB respectively (g TSS l-1), and 𝜃i an is the HRT of UASB (d).
Considering the UASB unit as an anaerobic process, Eq. (6) with hydrolysis
rate coefficient and Eq. (7) with hydrolyzed substrate transport rate coefficient will
be linearized to find Eq. (8).
−𝑑𝑆h
= 𝐾h (𝑆an − 𝑆h ) (6)
𝑑𝑡
−𝑑𝑆h
= 𝑘𝑆h 𝑋an (7)
𝑑𝑡
an𝐾 𝑆
𝑆h = 𝑘𝑋 h +𝐾 (8)
an h
Where 𝐾h is the hydrolysis rate coefficient (d−1) and 𝑘 is the hydrolyzed substrate
transport rate coefficient (L g-1 d-1).
Then, Eq. (1), Eq. (8), and Eq. (5) are combined to form Eq. (9). The Eq. (9)
was then modified and combined with Eq. (3) to form the final equation for effluent
estimation of the UASB unit (𝑆an Est). The detailed derivation of the 𝑆an Est
equations is detailed in Appendix 4.
𝜇max 𝐾 (𝑘(𝑌 (𝑆in −𝑆an )+𝑋an eff )+𝐾h )
= s +1 (9)
𝜇 𝐾 𝑆 h an
𝑘𝑌 𝑆in + 𝑘𝑋an eff + 𝐾h
𝑆an (Est) = (10)
𝐾h 𝜇max
𝑋an (t1) −1 +𝑘𝑌
𝐾s
ln
𝑋an (t0) 𝑋an eff
+𝐾e +
( (𝑡1 −𝑡0 ) 𝑋an (t1) 𝜃i an )
The effluent testing results in the UASB unit were used to create effluent
estimation models for 30 days from the last three variations. The effluent estimation
and the biokinetic value of the Monod Model were obtained from VBA Solver
coding in Microsoft Excel 2019 using the coding in Appendix 5. The VBA Solver
coding was modified from the previous research of Amelia (2022).
Firstly, the initial measurement data such as 𝑆an , 𝑆in , 𝑋an eff , 𝑋an , 𝜃i an , and t
were inputted into Excel to obtain an estimation model of UASB effluent.
10
Besides the Monod Biokinetic Model as the UASB unit effluent estimation
model, the Artificial Neural Network (ANN) was also programmed to predict the
effluent concentration of UASB using VBA Solver in Microsoft Excel. The ANN
operated in the backpropagation program with VBA coding in Appendix 6 from the
ANN coding of Arif (2021). The backpropagation program is a multilayer learning
algorithm to minimize errors by adjusting the weights based on the difference in
output and the desired target (Supriyanto et al. 2022). The ANN for UASB effluent
11
estimation used three layers of the calculation process, which are the input layer,
hidden layer, and output layer. Figure 2.6 presented the formation of four nodes of
the input layer, ten nodes of hidden layers, and one node of the output layer were
used to obtain the effluent estimation value. The number of nodes in the input and
output layers was adjusted to the availability of input data and the desired output
data. However, to produce high-accuracy forecasts in this study, more than one
hidden layer node was used in the model (Suhermi et al. 2018). Therefore, ten nodes
of the hidden layer were chosen to estimate the effluent of UASB in this research.
The steps involved in an ANN process are divided into the initiation, the
feedforward & backpropagation training, and the model fine-tuning process.
Firstly, random weights and biases were initiated to avoid networks stuck using
random values from the normal or uniform distribution. Then, the training process
was conducted to find the best function for the hidden layer and to create the
accurate estimation for the output layer (Fard et al. 2020). The training parameter
of 𝑆in , 𝑋an , 𝑋an eff, and 𝜃i an inserted in the ANN program as the input parameters,
while the 𝑆an inserted as the output parameter. The training data for each parameter
is attached in Appendix 7. The training process was calculated using feedforward
propagation to calculate the output target using the weights, biases, and activation
function. In the feedforward propagation, each input parameters forwarded signals
to all hidden layer unit. Then, the hidden layer (hi) accumulated all the input’s signal
using Eq. (11). The hidden layer obtained by combining the Eq. (11) with Sigmoid
Function in Eq. (12). Then, Eq. (13) used to sum up the weight input signals.
ℎinj = ∑𝑛𝑖=0 𝑋t ∙ 𝑉ij (11)
1
ℎinj = 𝑓 (ℎinj ) = −ℎin (12)
1+𝑒 j
ℎinj = ∑𝑛𝑖=0 𝐻t ∙ 𝑊ij (13)
Furthermore, the backpropagation algorithm was used to compute the error
gradients and adjust the weights and biases in the network using Eq. (14). The
hidden layer and output layer was calculated to determine the value of the weighted
correction (Wj) using Eq. (15) and Eq. (16).
1
Error = 2 (𝑂t − 𝑌t )2 (14)
∆𝑊j = 𝜂 ∙ 𝛿i ∙ 𝑋i (15)
𝛿k = (𝑂t − 𝑌t ) ∙ 𝑌 ∙ (1 − 𝑌t ) (16)
12
Improved weighting (Vij) was determined from the value of the hidden layer
with the input layer using the Eq. (17) and Eq. (18). New weights from Eq. (19) and
Eq. (20) was then obtained with the influence of old weights.
∆𝑉ij = 𝜂 ∙ 𝛿j ∙ 𝑋i (17)
𝛿j = 𝐻j ∙ (1 − 𝐻j ) ∙ ∑ 𝑊j ∙ 𝛿𝑘 (18)
𝑊j (new) = 𝑊j (old) + ∆𝑊j (19)
𝑉ij (new) = 𝑉ij (old) + ∆𝑉ij (20)
During training, model Fine-Tuning also proceeded to optimized the
parameter to find weight and bias values that produce network outputs that closely
match the targets. Then, the network’s performance is evaluated using validation
data to monitor progress and prevent overfitting. If performance meets
requirements, the network can be used for data modeling and prediction by inserting
the input values into the program.
The statistical analysis for the effluent estimation of UASB was evaluated
using the Analysis of Variance (Anova), t-Test, and the Mean Average Percentage
Error (MAPE). The Anova was carried out for effluent concentration data and
effluent estimation model data using Data Analysis of Anova : Single Factor in
Microsoft Excel 2019. However, one way Anova could also be used to assess the
effects of the HRT changes on effluent concentrations (Putri et al. 2021). The t-Test
was conducted to compare the effluent concentration value of the UASB with the
estimation value using the Data Analysis of t-Test: Paired Two Sample for Means
in Microsoft Excel 2019. The t-Test is suitable for the test for effluent value with
independent one-treatment method data (Bourget 2023). The statistic analysis was
also conducted using MAPE for the UASB unit’s effluent estimation models to
choose the best model between the Monod Model & the ANN. The MAPE was
carried out to investigate the accuracy of the model using the absolute average from
the difference of effluent estimation and effluent experiment data. The accepted and
the best model is chosen by the lowest value of MAPE (Maziya et al. 2016). The
MAPE equation is presented in Eq. (11).
1 𝐴 −𝐹
MAPE = 𝑛 ∑𝑛𝑡=1 | t𝐴 t| (11)
t
Where, 𝐴t is the actual value from the laboratory measurement data and 𝐹t is the
Forecast value from the estimation models.
Figure 2.7 Goldsim Pro 14.0 properties interface for sensitivity analysis
13
valves for supernatant and biomass, sludge controlling valves, and a gas outlet
valve. The UASB unit is divided into three distinct zones: digestion, transition, and
settling zone (Rajagopal et al. 2019). Firstly, POME enters through the inlet at the
bottom of the unit in the digestion zone. The POME then uniformly flows upward
between the anaerobic sludge blanket that contains the flocculent sludge in the
digestion zone. Anaerobic decomposition occurs due to the digestion of POME
dissolved substrate by the sludge (Elmitwalli 2000). The POME then flows into the
transition zone. While most of the sludge remains in the digestion zone, some
sludge particles are carried upward. The settling zone allows the sludge particles to
settle back into the digestion zone while the liquid effluent flows out to the IFAS
unit (Hickey et al. 1991; Lettinga and Hulshoff Pol 1991; Tauseef et al. 2013).
(a) (b)
Figure 3.1 Design of UASB unit (a) front view (b) left side view
Following the treatment process in the UASB, the partially treated POME
was directed to the IFAS unit, which incorporates a combination of attached and
suspended growth biomass. Subsequently, the clarifier unit settled the suspended
solid, including biomass that may have been washed out from the system, allowing
the clarified effluent to be discharged (Qasim and Zhu 2018). The biomass is then
transferred to the stabilization unit, where the microorganisms present in the
biomass experience a period of nutrient starvation due to the absence of substrate.
The biomass was subsequently recirculated back to the IFAS unit under these
starving conditions, expecting to exhibit increased capability for organic matter
reduction (Amelia 2022).
16
value indicates the healthy growth of microorganisms during the seeding and
acclimatization process, as TSS serves as an estimation approach to assess
microorganism concentrations involved in substrate degradation. The seeding and
acclimatization process is considered complete when biomass concentrations
exceed 4000 mg/L (Syahrin et al. 2016).
Figure 3.3 compare the pH data of the influent and effluent of the ACSt
reactor with allowable pH standard for POME from Indonesian Ministry of
Environment Regulation No. 5/2014. Based on the graphic, the pH of raw POME
in the influent of ACSt reactor was relatively low with pH 3-5 range, which does
not meet the standard pH range of 6-9. However, the effluent of the ACSt reactor
falls around the pH range of 6-9. Notably, only the first and sixth variations of the
HRT achieve this standard. Moreover, the fluctuation of pH in the graphic formed
a trendline of decreased pH during the first four HRT variations. However, the last
three HRT variations formed an increased trendline of pH. This result was due to
the HRT change of units in ACSt reactors.
Figure 3.4 compared the TSS data of influent and effluent of ACSt reactor
with the standard value of 250 mg/L from Indonesian Ministry of Environment
18
Regulation No. 5/2014. The influent of ACSt reactors also has a high amount of
TSS, which was successfully degraded into the standard value of 250 mg/L by the
Reactor in the first third HRT variations. While for the other variations, the effluent
TSS value highly fluctuated above 250 mg/L as the HRT variations became faster
for each unit. This result caused by the utilization time of microorganisms became
lesser. The high TSS content in POME is attributed to the presence of fine fruit
fibers that pass through the process and become flow to the wastewater as
suspended solids (Wahyuni 2017). Moreover, the CODs value of POME in the
influent and effluent of the ACSt reactor was compared in Figure 3.5 with the
standard value of 350 mg/L from Indonesian Ministry of Environment Regulation
No. 5/2014. The graph reveals that only the first and second HRT variations were
able to effectively degrade the CODs below the standard value of 350 mg/L. The
five other HRT variations fluctuated above the standard criteria of 350 mg/L CODs.
Figure 3.4 The TSS data of influent and effluent of ACSt reactor
Figure 3.5 The CODs data of influent and effluent of ACSt reactor
19
Based on the comparison of the measured pH, TSS, and CODs value with the
effluent standard, the first variation with the HRT of 8, 0.67, and 0.25 days for
UASB, IFAS, and Stabilization unit, respectively, proves to be the most successful
in achieving all the standard criteria for POME treatment. This result was attributed
to the longest HRT duration from the variations of this research. The low HRT
resulted in limited contact time between the biomass and the substrate, reducing the
opportunity for substrate degradation. The result was connected with the research
from Rekoyoso et al. (2014), which stated that the prolonged duration of HRT
results in an increased contact time between microorganisms and the substrate.
Consequently, substrate utilization time is extended and gives a greater removal
efficiency.
However, even after proving to be the most successful variation in meeting
all the standard criteria for POME treatment, the first variation of ACSt was not the
best in the removal efficiency for any parameter, as presented in Table 3.2. The best
average removal of TDS and ammonia parameters were the third variation, with
88.50% TDS removal and 99.89% ammonia removal. Then, the best average
removal of the TSS parameter was the second variation of 99.67% TSS removal.
Lastly, 99.88% of COD’s average removal was obtained from the fifth HRT
variation. This result is due to the fluctuation of Influent and effluent of POME in
the ACSt reactor, which caused the average removal for each HRT variation to
become not representative. By then, the average removal efficiency data of the
ACSt reactor cannot be used before the data is smoothed or more data sets are
obtained.
Table 3.2 Effluent measurement and average removal of the ACSt reactor
Varia- Effluent Range Average Removal (%)
tion pH Temp (oC) TDS TSS COD Ammonia OG TN
1 7.16-8.21 22.8-31.1 69.39 99.38 99.62 83.68
2 5.74-8.35 25.9-30.8 65.04 99.67 99.81 78.62
3 5.23-8.05 26.6-30.6 88.50 99.53 99.26 99.89
4 4.83-7.19 26.6-31.4 78.06 97.19 99.72 33.50 99.74 96.73
5 5.09-7.88 22.9-31.9 69.78 96.86 99.88 77.17
6 7.16-8.28 24.5-32.4 41.83 99.34 95.73 99.78
7 7.93-9.18 25.7-29.7 26.63 98.93 94.96 93.29
the OG value significantly decreases by 99.74% and complies with the standards,
reaching a value of 10.99 mg/L. At the same time, the TN value was 67.5 mg/L,
slightly above the allowable standard of 50 mg/L by the Indonesian Ministry of
Environment Regulation No. 5/2014. This TN value then became 2.21 mg/L after
being processed in the ACSt reactor with the removal of 96.73%. The total nitrogen
represented the total amount of nitrogen in the water and not in the microorganisms,
which include nitrate, ammonia, free amino acids, and other macromolecules of
dissolved organic nitrogen (Broch 2015).
3.4 The UASB Unit Performance Evaluation
As the first unit of the ACSt reactor, the UASB unit plays a crucial role in
significantly degrading high-strength pollutants present in the POME. In order to
assess the degradation efficiency of the UASB on 8, 4, and 2 days HRT variations,
the POME treatment had measured for the 40th until 70th days of the experiment
with ten days each for every HRT variation. The measurement results for the
influent and effluent of the UASB for pH parameter in Figure 3.6 seems to have
insignificant change with the acidic range of pH 3.44 – 4.93 due to the uncontrolled
pH conditioning. These pH values do not achieve the effluent pH standard of pH 6-
9 for POME by the Indonesian Ministry of Environment Regulation No. 5/2014.
Furthermore, the trend of pH is not significantly different from each HRT variation,
which explains that HRT change in UASB did not influence the pH change in the
treatment process of POME.
As a consequence of the acidic pH conditions, methane gas production was
not measured in this research, as methanogenic bacteria are highly sensitive to pH
changes and require an optimum pH range of 6.5-7.5 for their activity (Ikbal and
Nugroho 2006). Thus, it can be assumed that the reactor's bacteria are
predominantly acid-forming rather than methane-producing bacteria, as the former
has a wider pH tolerance range compared to the latter. With the presence of acid-
forming bacteria, acidogenesis reactions occur within the UASB unit. This
acidogenesis reaction is characterized by the high acidity in the reactor, as indicated
by the pH data (Rambe 2016).
Ammonia concentrations of both influent and effluent of the UASB unit were
also shown in Figure 3.6, with the range of 5.75-80.56 mg/L. Based on the graphic,
the UASB’s effluent ammonia value changed with the HRT variations in the
reactor. In this case, the HRT of two days has the lowest overall ammonia value,
while the HRT of four days has the highest value. The nonlinear correlation between
effluent ammonia in UASB and HRT changes resulted from ammonia fluctuation
in POME. The concentration of ammonia in wastewater can influence the ability of
microorganisms to degrade organic matter (Indradewi et al. 2015). Ammonia
concentrations ranging from 50-200 mg/l are supportive to microorganisms,
concentrations of 200-1000 mg/l do not have an impact on microorganisms,
concentrations of 1500-3000 mg/l act as inhibitors or impair the performance of
microorganisms at specific pH levels, and concentrations exceeding 3000 mg/L are
toxic to microorganisms (Rajagopal et al. 2013). In this study’s case of the UASB
unit, the range of ammonia concentrations observed 5.75-80.56 mg/L falls within
the supportive range for the microorganisms’ activity.
Figure 3.7 The TDS and temperature measurement data of UASB Unit
Figure 3.7 presented the temperature value ranging from 20.1 to 33.4 oC and
the TDS value ranging from 131 to 2490 ppm. The measurement result for the
influent and effluent of the UASB for temperature parameter indicate insignificant
changes, which can be attributed to the ambient atmospheric conditions. Anaerobic
treatment under mesophilic microorganism conditions can be effectively conducted
at ambient temperatures without needing external heat input in tropical regions
where temperatures are relatively stable and range from 28 to 33 ℃, like Indonesia
(Ikbal and Nugroho 2006).
The TSS value from influent and effluent of the UASB exhibit a significant
reduction, decreasing from 1800-98200 mg/L to 143-6600 mg/L, as shown in
Figure 3.8. This substantial degradation can be attributed to the upflow concept of
UASB that settles most of POME’s suspended solid. Consequently, the sludge bed
increases within the UASB unit (Utami et al. 2016). Regarding the CODs value in
Figure 3.9, a significant removal was observed from the range of 15592-39075
mg/L to 1553-24688 mg/L. However, both TSS and CODs parameters from
UASB’s influent and effluent have not achieved the standard value of POME from
the Indonesian Ministry of Environment Regulation No. 5/2014, which are 250
mg/L for TSS and 350 mg/L for CODs.
Table 3.3 provides the removal efficiencies of CODs and TSS, indicating that
the two days HRT variation demonstrates the highest removal efficiency, with 68%
for CODs and 93% for TSS in POME. Additionally, the two days HRT has the
smallest variance (161.9) compared to the other HRT values. However, it is
23
important to note that the CODs and TSS values from the UASB unit still need to
achieve the effluent standard regulation in Indonesia. Further processing within the
IFAS, sedimentation, and stabilization units on the ACSt reactor, as described in
the previous subsection, aims to further treat the effluent from UASB to comply
with the standard value of POME from the Indonesian Ministry of Environment
Regulation No. 5/2014 before being released into the environment.
3.5 The Effluent Concentration Estimation Modeling
The estimation of effluent concentrations in the UASB unit under unsteady
state conditions was performed using the Monod Model and the ANN. The graphic
from Figure 3.10 and the table from Appendix 8 compared these estimation models
with laboratory experiment data, revealing a fluctuating result of measurement and
model. However, the experiment data is closer to the ANN model estimation than
the Monod Model. This result can be attributed to the capability of ANN to predict
complex & non-linear relationships and create a generalized structure for all
datasets, unlike conventional models such as the Monod Model (Elnekave et al.
2012). Additionally, the Monod Model predicts effluent concentration without
considering the presence of inhibitors, resulting in a highly simplified estimation.
Statistical analysis had to be conducted to validate this hypothesis.
Figure 3.10 The comparison of experiment data, Monod Model, and ANN
value from Lewis (1982) in Table 3.4, the Monod Model is reasonable to be used
for effluent estimation in the UASB unit. Nevertheless, the MAPE value from the
ANN model appears to be more promising for effluent estimation in UASB unit
due to the well accuracy level. Thus, the results of ANN effluent estimation on the
UASB unit provide sufficiently precise results with laboratory experiment results.
In future field-scale processing of UASB on POME, the trained ANN model from
this research could be applied to predict substrate concentrations in the effluent.
While ANN has demonstrated accurate estimations, the Monod Model remains
necessary to obtain estimations of biokinetic coefficients.
Table 3.4 The interpretation of MAPE value
MAPE Value (%) Interpretation
<10 Highly accurate
10-20 Moderate
20-50 Reasonable
>50 Inaccurate
used the UASB unit as treatment process for the POME, similar to this research.
The 𝑌 value represents the capability of the microorganism to synthesize new cells
for biomass growth. The 𝑌 value of 0.98 mg TSS/mg CODs was also inside the
range of 0.052 until 3.906 mg TSS/mg CODs from Table 3.6.
The 𝐾h value of 0.58 d-1 was also inside the range of 0.21-0.66 d-1 from the
research of Nakhla et al. (2006). This value represented substrate transports as the
intracellular and extracellular hydrolysis process accumulated inside the
microorganism, which also acted as rate-limiting phase during the anaerobic
biodegradation of particulate matter (Kurniawan et al. 2021). Lastly, the 𝑘 value of
0.0004 d-1 represents a very small value of the rate of substrate transport inside the
biomass. Overall, the biokinetic parameters obtained from the Monod Model
provide valuable information about the efficiency of substrate degradation and
microbial growth within the UASB unit. These parameters are crucial for
understanding and optimizing the performance of the reactor in treating POME.
Table 3.6 The reference biokinetic parameter
Unit Wastewater 𝑲𝐬 𝝁𝐦𝐚𝐱 𝑲𝐞 𝒀 Ref.
UASB POME 3459 1.861 0.233 3.906 (Ibrahim et al. 2021)
UASFF POME 6000 0.568 0.002 0.786
(Akhbari et al. 2020)
UASB COD 3180 0.840 0.009 0.125
UASFF POME 982 0.207 - 0.174 (Zinatizadeh et al. 2006)
UASB Syntetic 560 0.213 0.093 0.780 (Sponza and Uluköy 2008)
UASB Textile 4000 0.105 0.006 0.125 (Isik and Sponza 2005)
CD POME 270 0.117 0.020 0.052 (Yeoh 1986)
IAAB POME 8168 0.103 0.038 0.196 (Chan et al. 2017)
ABSR POME 203 0.524 0.024 0.990 (Wong et al. 2009)
MABR POME 313 0.304 - - (Faisal et al. 2001)
Min 203 0.103 0.002 0.0520
Max 8168 1.861 0.233 3.9060
Indeed, the biokinetic parameters of the UASB unit influence the effluent
estimation using Monod Model. Sensitivity analysis is performed to evaluate the
influence of each biokinetic parameter on the effluent estimation of the UASB unit.
The largest significant influence of each biokinetic parameter was from the 𝜇𝑚𝑎𝑥
with 29,55%. Followed by the 𝐾s , 𝑘, 𝐾h , 𝑌, and 𝐾e with the value of 19.38%,
16.39%, 16.39%, 11.04%, and 7.26% (Figure 3.11). The largest significance by
𝜇𝑚𝑎𝑥 was caused by the change of microorganisms value as the main processing
component of POME in the UASB. This result was aligned with the Figure 3.12
(b), which presented a largest change in effluent estimation (San) for up to 1262
mg/L, due to the change of 𝜇𝑚𝑎𝑥 value into 1.698 and 2.076 d-1 as the ± 10% values
of selected 𝜇𝑚𝑎𝑥 (1.887 d-1). Moreover, the second largest significance was due to
the 𝐾s parameter. It obtained effluent estimation (San) change for up to 1007 mg/L
by the change of 𝐾s value into 3113 and 3805 mg CODs/L as the ± 10% values of
selected 𝐾s (1.887 mg CODs/L) (Figure 3.12 (c)).
In contrast, the smallest significance of the Ke parameter indicates that
microorganism decay does not substantially impact the effluent estimation in the
same magnitude as the microbial maximum growth rate. This result was also
aligned with the Figure 3.12 (a), which presented a smallest change in San to the
26
maximum value of 570 mg/L, due to the change of Ke value into 0.1434 and 0.1752
d-1 as the ±10% values of Ke (0.1593 d-1). These findings align with the study
conducted by Mardani et al. (2011), which revealed that the parameter 𝜇max is more
sensitive to effluent concentration compared to 𝐾s and 𝐾e . This sensitivity analysis
highlights the importance of accurately determining and estimating the biokinetic
parameters, particularly the 𝜇max and 𝐾s values, as they have the greatest influence
on the effluent estimation of the UASB unit. Proper calibration and optimization of
these parameters can enhance the accuracy and reliability of the Monod Model for
predicting effluent concentrations in the UASB unit treating POME.
The biokinetic parameters of the Monod Model have various effects on the
effluent estimation in UASB. The graphics in Figure 3.12 display the influence of
each biokinetic parameter on the estimation of 𝑆an for HRT variations from 0,5 -
15 days. Based on the graph, the 𝑆an estimation value tends to decrease during the
longer HRT days of the UASB unit. However, beyond an HRT of more than two
days, the slope of the 𝑆an estimation in the graphic becomes flatter. This result
aligns with the previous finding that two days HRT was the most efficient for
POME degradation in the UASB unit. The longest HRT duration is not feasible for
real-scale wastewater treatment due to the high volume of wastewater that needs to
be treated. Thus, the HRT of two days is the most efficient HRT to degrade POME
with the most efficient time consideration.
Furthermore, Figure 3.12 also displays the effect of biokinetic value on the
effluent estimation. The value of 𝐾e and 𝜇max from Figure 3.12 (a) and Figure 3.12
(b) are inversely related. A higher 𝐾e value leads to higher effluent estimation,
whereas a higher 𝜇max value results in lower effluent estimation. Consequently, the
Ke value represents the endogenous decay of microorganisms, while the 𝜇max value
represents their growth. Similarly, the value of 𝜇max and 𝐾s from Figure 3.12 (b)
and Figure 3.12 (c) also has an inverse correlation. The higher the 𝐾s is, the lower
the 𝜇max will be. Moreover, the biological treatment efficiency will decrease and
27
vice versa (Mousavian et al. 2019). The 𝑌 from Figure 3.12 (d) estimates the total
of sludge produced inside the unit (Enitan and Adeyemo 2014). Therefore, the
bigger 𝑌 value results in less efficiency in the treatment. The value of 𝐾h and 𝑘
from Figure 3.12 (e) and Figure 3.12 (f) also has an inverse relation. When the 𝐾h
value is high, the substrate removal efficiency is also high. However, when the 𝑘
value is high, the substrate processing efficiency becomes low because the 𝐾h
coefficient represents the hydrolysis rate coefficient, which acts as the limiting rate
for anaerobic biodegradation processes. In contrast, the 𝑘 coefficient represents the
transport rate coefficient of substrate hydrolysis (Ebner et al. 2016).
standards and safely discharge the POME’s effluent into the environment. Based
on this research, the ACSt reactor can be scaled up and utilized for on-field POME
treatment. However, further research is needed to adapt and optimize the
application of this technology on a larger field scale. Moreover, these findings
provide valuable insights for the design and operation of ACSt reactor in general
and the UASB Unit in particular for treating POME, potentially leading to improved
wastewater treatment efficiency in the industry. The research also introduces ANN
as a tool for estimating effluent from the UASB unit. Manual performance
measurement and estimation efforts can be reduced significantly by utilizing ANN.
The result can lead to more efficient monitoring and control of the UASB unit,
saving labor and resources. Estimations for biokinetic parameters of the UASB unit
under unsteady-state conditions were also obtained. These parameters, such as 𝐾e ,
𝜇𝑚𝑎𝑥 , 𝐾s , 𝑌, 𝐾h , and 𝑘, are crucial for understanding the microbial processes
involved in treating POME. Understanding the effects of these parameters allows
for better optimization and control of the UASB system to achieve desired treatment
outcomes for POME. In summary, this research contributes to advancing
sustainable wastewater treatment practices in the Palm Mill Oil industry and
provides practical solutions to address the challenges of treating high-contaminant
effluents. The findings can have real-world applications in improving treatment
efficiency, reducing manual efforts, and promoting environmentally friendly
practices in the industry.
29
CONCLUSION
REFERENCES
doi:10.1080/09593330.2016.1217053.
Chan YJ, Tan WJR, How BS, Lee JJ, Lau VY. 2015. Fuzzy optimisation approach
on the treatment of palm oil mill effluent (POME) via up-flow anaerobic
sludge blanket–hollow centered packed bed (UASB–HCPB) reactor. J Water
Process Eng. 5:112–117. doi:10.1016/J.JWPE.2015.01.005.
Chong S, Sen TK, Kayaalp A, Ang HM. 2012. The performance enhancements of
upflow anaerobic sludge blanket (UASB) reactors for domestic sludge
treatment – A State-of-the-art review. Water Res. 46(11):3434–3470.
doi:10.1016/J.WATRES.2012.03.066.
Chung CY, Selvarajoo A, Sethu V, Koyande AK, Arputhan A, Lim ZC. 2018.
Treatment of Palm Oil Mill Effluent (POME) by coagulation flocculation
process using peanut–okra and wheat germ–okra. Clean Technol Environ
Policy. 20(9):1951–1970. doi:10.1007/S10098-018-1619-Y/METRICS.
Ebner JH, Labatut RA, Lodge JS, Williamson AA, Trabold TA. 2016. Anaerobic
co-digestion of commercial food waste and dairy manure: Characterizing
biochemical parameters and synergistic effects. Waste Manag. 52:286–294.
doi:10.1016/j.wasman.2016.03.046.
Elmitwalli TA. 2000. Anaerobic Treatment of Domestic Sewage at Low
Temperature. Wageningen: Wageningen University.
Elnekave M, Celik SO, Tatlier M, Tufekci N. 2012. Artificial neural network
predictions of up-flow anaerobic sludge blanket (UASB) reactor performance
in the treatment of citrus juice wastewater. Polish J Environ Stud. 21(1):49–
56.
Enitan AM, Adeyemo J. 2014. Estimation of bio-kinetic coefficients for treatment
of brewery wastewater. Int J Environ Ecol Eng. 8(6):527–531.
Faekah IN, Fatihah S, Mohamed ZS. 2020. Kinetic evaluation of a partially packed
upflow anaerobic fixed film reactor treating low-strength synthetic rubber
wastewater. Heliyon. 6(3):e03594. doi:10.1016/J.HELIYON.2020.E03594.
Fairuzah N, Asnawi M. 2012. Biological nutrient removal of Palm Oil Mill Effluent
(POME) Hybrid Sequencing Batch Reactor (H-SBR). Johor Bahru: Universiti
Teknologi Malaysia.
Faisal M, Unno, Hajime. 2001. Kinetic analysis of palm oil mill wastewater
treatment by a modified anaerobic baffled reactor. Biochem Eng J. 9:25–31.
Fard MB, Mirbagheri SA, Pendashteh A, Alavi J. 2020. Estimation of effluent
parameters of slaughterhouse wastewater treatment with artificial neural
network and B-spline quasi interpolation. Int J Environ Res. 14(5):527–539.
doi:10.1007/S41742-020-00274-1/METRICS.
Gutiérrez S, Hernández A, Viñas M. 1999. Mechanism of degradation of wool wax
in the anaerobic treatment of woolscouring wastewater. Water Sci Technol.
40(8):17–23. doi:10.1016/S0273-1223(99)00604-6.
Harmawan T. 2022. Analisis Kandungan Minyak dan Lemak pada Limbah Outlet
Pabrik Kelapa Sawit di Aceh Tamiang. Quim J Kim Sains dan Terap. 4(1):15–
19. doi:10.33059/jq.v4i1.4318.
Hickey RF, Wu WM, Veiga MC, Jones R. 1991. Start-up, operation, monitoring
and control of high-rate anaerobic treatment systems. Water Sci Technol
Technol. 24:207–255.
Ibrahim MM, Jemaat Z, Nour AH. 2021. Performance and kinetic evaluation of
Palm Oil Mill Effluent (POME) digestion in a continuous high rate Up-Flow
32
APPENDIX
38
Monod Equation
𝜇max 𝑆h
𝜇= (5)
𝐾s + 𝑆h
𝑆h (𝑘𝑋an + 𝐾h ) = 𝐾h 𝑆an
𝐾h 𝑆an
𝑆h = (8)
𝑘𝑋an +𝐾h
Sub Run_Monod_UASB()
Dim a As Integer
For a = 0 To 28 ' the total data
SolverReset
'SolverOptions Iterations:=10000, Precision:=0.001
SolverOk SetCell:="$X$" & (5 + a), MaxMinVal:=2, ByChange:="$M$" & 7 & ",$N$" & 7 & ",$O$" & 7 &
",$P$" & 7 & ",$Q$" & 7 & ",$R$" & 7
SolverAdd CellRef:="$M$" & 7, relation:=3, FormulaText:="$M$4" 'Ke min
SolverAdd CellRef:="$M$" & 7, relation:=1, FormulaText:="$M$5" 'Ke max
SolverAdd CellRef:="$N$" & 7, relation:=3, FormulaText:="$N$4" 'myumax min
SolverAdd CellRef:="$N$" & 7, relation:=1, FormulaText:="$N$5" 'myumax max
SolverAdd CellRef:="$O$" & 7, relation:=3, FormulaText:="$O$4" 'Ks min
SolverAdd CellRef:="$O$" & 7, relation:=1, FormulaText:="$O$5" 'Ks max
SolverAdd CellRef:="$P$" & 7, relation:=3, FormulaText:="$P$4" 'Y min
SolverAdd CellRef:="$P$" & 7, relation:=1, FormulaText:="$P$5" 'Y max
SolverAdd CellRef:="$Q$" & 7, relation:=3, FormulaText:="$Q$4" 'Kh min
SolverAdd CellRef:="$Q$" & 7, relation:=1, FormulaText:="$Q$5" 'Kh max
SolverAdd CellRef:="$R$" & 7, relation:=3, FormulaText:="$R$4" 'k min
'SolverAdd CellRef:="$R$" & 7, relation:=1, FormulaText:="$R$5" 'k max
SolverSolve True
'copy all the temporary value to the next table for each experiment day
Cells(5 + a, 27) = Cells(7, 13)
Cells(5 + a, 28) = Cells(7, 14)
Cells(5 + a, 29) = Cells(7, 15)
Cells(5 + a, 30) = Cells(7, 16)
Cells(5 + a, 31) = Cells(7, 17)
Cells(5 + a, 32) = Cells(7, 18)
Cells(5 + a, 33) = Cells(7, 19)
Next a
MsgBox "Done Monod"
End Sub
Sub Run_SSD_UASB()
Dim a As Integer
For a = 0 To 28
Cells(7, 36) = Cells(5 + a, 27)
Cells(7, 37) = Cells(5 + a, 28)
Cells(7, 38) = Cells(5 + a, 29)
Cells(7, 39) = Cells(5 + a, 30)
Cells(7, 40) = Cells(5 + a, 31)
Cells(7, 41) = Cells(5 + a, 32)
SolverReset
50
For jj = 1 To ni
xinp(jj, ii) = Sheet6.Cells(ii + 6, jj + 1 + no)
Next jj
Next ii
End Sub
Rem Data Identification
Sub DataIdentification()
For i = 1 To nl
nn(1) = ni
nn(2) = nh
nn(3) = no
Next i
For ii = 1 To nn(1)
mixi(ii) = xinp(ii, 1)
Next ii
For ii = 1 To nn(1)
For jj = 1 To ncase
If xinp(ii, jj) < mixi(ii) Then mixi(ii) = xinp(ii, jj)
Next jj
Next ii
For ii = 1 To nn(1)
For jj = 1 To ncase
If xinp(ii, jj) > maxi(ii) Then maxi(ii) = xinp(ii, jj)
Next jj
Next ii
For ii = 1 To nn(nl)
For jj = 1 To ncase
If des(ii, jj) > maxo(ii) Then maxo(ii) = des(ii, jj)
Next jj
Next ii
For ii = 1 To nn(nl)
mixo(ii) = 100000
For jj = 1 To ncase
If des(ii, jj) < mixo(ii) Then mixo(ii) = des(ii, jj)
Next jj
Next ii
End Sub
Sub Desimalisasi()
For ii = 1 To nn(1)
For jj = 1 To ncase
xinp(ii, jj) = (xinp(ii, jj) - mixi(ii)) / (maxi(ii) - mixi(ii))
Next jj
Next ii
For ii = 1 To nn(nl)
For jj = 1 To ncase
des(ii, jj) = ((des(ii, jj) - mixo(ii)) * 0.6 / (maxo(ii) - mixo(ii))) + 0.2
Next jj
Next ii
End Sub
Sub Result()
For ii = 0 To (nn(nl) - 1)
For jj = 1 To ncase
Sheet5.Cells(2 + jj, 5) = jj
Sheet5.Cells(2 + jj, 6 + (ii * 2)) = Sheet6.Cells(jj + 6, 2 + ii)
Next jj
Next ii
End Sub
Sub TrainingProcess()
r = Val(Right$(Time$, 2))
Randomize r
53
TW = 1
For ii = 1 To nl
For kk = 1 To nn(ii)
For jj = 1 To nn(ii - 1)
w(ii - 1, jj, kk) = 2 * Rnd - 1
TW = TW + 1
Next jj
Next kk
Next ii
'Start Iteration Process
itend = iterasi
ite = 0
For iii = 1 To itend
erss = 0
For ii = 2 To nl
For kk = 1 To nn(ii)
For jj = 1 To nn(ii - 1)
dwe(ii - 1, jj, kk) = 0
dwes(ii - 1, jj, kk) = 0
dw2s(ii - 1, jj, kk) = 0
dw2(ii - 1, jj, kk) = 0
Next jj
Next kk
Next ii
For i = 1 To ncase
For jj = 1 To nn(1)
X(1, jj) = xinp(jj, i)
Next jj
'forward (calculating output of each PE)
For ii = 2 To nl
For kk = 1 To nn(ii)
u(ii, kk) = 0
For jj = 1 To nn(ii - 1)
u(ii, kk) = u(ii, kk) + w(ii - 1, jj, kk) * X(ii - 1, jj)
Next jj
X(ii, kk) = 1 / (1 + Temp * Exp(-u(ii, kk)))
Next kk
Next ii
'Backward of nlth layer
For ii = 1 To nn(nl)
Err2 = des(ii, i) - X(nl, ii)
d(nl - 1, ii) = Err2 * Temp * Exp(-u(nl, ii)) / (1 + Exp(-u(nl, ii))) ^ 2
If Err2 > ep Then ep = Abs(Err2)
Next ii
'Backward for hidden layer
For ii = nl - 2 To 1 Step -1
For kk = 1 To nn(ii + 1)
d(ii, kk) = 0
For jj = 1 To nn(ii + 2)
d(ii, kk) = d(ii, kk) + w(ii + 1, kk, jj) * d(ii + 1, jj)
Next jj
d(ii, kk) = d(ii, kk) * Temp * Exp(-u(ii + 1, kk)) / (1 + Exp(-u(ii + 1, kk))) ^ 2
Next kk
Next ii
'summation of weight value error
For ii = 2 To nl
For kk = 1 To nn(ii)
For jj = 1 To nn(ii - 1)
dw(ii - 1, jj, kk) = d(ii - 1, kk) * X(ii - 1, jj)
dwe(ii - 1, jj, kk) = dwe(ii - 1, jj, kk) + dw(ii - 1, jj, kk)
54
Next jj
Next kk
Next ii
ersum(i) = 0
For ii = 0 To nn(nl) - 1
cal(ii + 1, i) = X(nl, ii + 1)
er(ii + 1, i) = des(ii + 1, i) - X(nl, ii + 1)
ersum(i) = ersum(i) + er(ii + 1, i) ^ 2
If iii = itend Then Sheet5.Cells(2 + i, 7 + (ii * 2)) = (X(nl, ii + 1) - 0.2) * (maxo(ii + 1) - mixo(ii + 1)) / 0.6
+ mixo(ii + 1)
Next ii
erss = erss + ersum(i)
Next i
'Correction of weight value
For ii = 2 To nl
For kk = 1 To nn(ii)
For jj = 1 To nn(ii - 1)
dwes(ii - 1, jj, kk) = dwes(ii - 1, jj, kk) + dwe(ii - 1, jj, kk)
dw2(ii - 1, jj, kk) = dwe(ii - 1, jj, kk)
Next jj
Next kk
Next ii
npemb = Int(ncase / 4)
For ii = 2 To nl
For kk = 1 To nn(ii)
For jj = 1 To nn(ii - 1)
w(ii - 1, jj, kk) = w(ii - 1, jj, kk) + Eta * dwes(ii - 1, jj, kk) / npemb + Alpha * dw2s(ii - 1, jj, kk) / npemb
dw2s(ii - 1, jj, kk) = dwes(ii - 1, jj, kk)
Next jj
Next kk
Next ii
erx = erss / nn(nl) / ncase
Delta = erx - ers
ers = erx
AdjAp = ep / 0.5
If AdjAp > 1 Then Temp = 1 / AdjAp 'adjusted gain parameter
If AdjAp <= 1 Then Temp = 1 'adjusted gain parameter
Sheet5.Cells(11, 2) = iii
Sheet5.Cells(12, 2) = Delta
Sheet5.Cells(13, 2) = ers
ite = ite + 1
ep = 0
Next iii
'Display
TW = 1
For ii = 2 To nl
For kk = 1 To nn(ii)
For jj = 1 To nn(ii - 1)
Sheet5.Cells(2 + TW, 10) = w(ii - 1, jj, kk)
TW = TW + 1
Next jj
Next kk
Next ii
End Sub
Sub Clear_ANN()
'delete all previous ANN value
Dim ANN_value As Range
Set ANN_value = Range("$F$3:$J$100")
ANN_value.ClearContents
End Sub
55
Appendix 8 The UASB’s effluent measurement data, Monod Model, and ANN
t Measurement Data (mg Monod Estimation (mg ANN Estimation
(d) sCOD/L) sCOD/L) (mg sCOD/L)
41 6580 5886 6186
42 18475 11198 16481
43 6375 12143 7126
44 9303 13147 7973
45 9107 9821 6855
46 24688 26879 24482
47 10039 13936 10718
48 9990 15055 12062
49 9352 10632 10135
50 9663 24817 10665
51 9523 14875 15110
52 19292 6436 17722
53 16962 5751 14631
54 23625 6762 17137
55 14551 7567 15639
56 14142 7084 12755
57 17902 15624 19015
58 16676 11776 16901
59 9483 8241 11528
60 6671 6641 9056
61 1553 11846 1355
62 5722 9532 9292
63 6458 5150 7812
64 7766 8809 6952
65 9360 11325 7109
66 4087 14036 4357
67 10463 16969 9827
68 10431 11098 11942
69 10039 9013 8282
70 9107 11323 7447
57
BIOGRAPHY
The author, Aliyah Baida Wiwiyanti, was born in Makassar, 21st of July
2001 as the only child of Aida Abbas. The author finished Senior High School in
SMA Negeri 9 Jeneponto on 2019. Then the author became an undergaduate student
in The Department of Civil and Environmental Engineering, Faculty of Agricultural
Engineering and Technology, IPB University on 2019 as an AdiK 3T Scholarship
holder. During four years in IPB, the author had the opportunities to attended the
short exchange course programme of Sakura Science Program in Kobe University,
Japan (2022) and the volunteering programme of Youth Agricareture: Zero Waste
in Bogor, Indonesia (2019).
The author had achieve the 1st Prize of Presentation and Poster Category on
the 35th Pekan Ilmiah Nasional (Pimnas) in PKM-KC entiteled “Rancang Bangun
Reaktor Kompak Berbasis Hybrid Attached Growth Biofilm Terintegrasi Arduino
Sebagai Inovasi Pengolahan Limbah Cair Tekstil” (2022); the Finalist of the 34th
Pimnas in PKM-GT entiteled “I-SURE: Smart Undersea Train for Indonesian
Unifier” (2021); the Recipient of Grant Fund Program Kreativitas Mahasiswa in
PKM-GFT entitled “D-TECSER: Integrasi Sistem Mitigasi Bencana
Hidrometeorologis Indonesia melalui Teknologi Modifikasi Cuaca Berbasis Drone
Laser dan Artificial Intelligence” and “I-FISHILLAGE: Integrated Fishing Village
for Robust Indonesian Fisheries System” (2022); the 1st Prize of Poster Design
Competition in “LISTEN” KN FGMMI (2021); and the 2nd Prize of Puteri OMDA
in Gebyar Nusantara in IPB University (2020).
Furthermore, the author also join several student organizations and
commitees. Some of them are DPM Fateta as general secretary (2022) & aspiration
commission secretary (2021), IGAF LC IPB as the member of MIC Division (2019-
2021), ICEF as the head of Publication Division (2022) & the staff of Sponsorship
Division (2021), PEMIRA-F as the head of Design and Documentation Division
(2021), The 2nd ISCEE as the staff of Design Division (2021), and MPKMB IPB
57 as the staff of Decoration, Design, and Branding (2020).
In addition, the author become the intern at the Leuwikeris Dam Project (1st
Batch) by PP-BBN KSO with the internship report entitled “Evaluasi Analisis
Mengenai Dampak Lingkungan (Amdal) pada Quarry Area Proyek Konstruksi
Bendungan Leuwikeris, Jawa Barat” (2023); the lecturer assistant in the courses of
“Ilmu Ukur Wilayah”, “Gambar Teknik Konstruksi”, “Mekanika Fluida dan
Hidrolika”, “Hidrologi Teknik”, and “Pengelolaan Limbah Cair” (2021-2023); and
the co-writer of scientific publication in Jurnal Teknik Sipil Institut Teknologi
Padang entitled “Efek Kecepatan Pengendapan terhadap Perencanaan Unit
Sedimentasi Primer pada Pengolahan Air Lindi” (2022).