Professional Documents
Culture Documents
Concept Note 1
Plenary Speakers 2
Technical Program (Energy Session) 4
Technical Program (Environment and Infrastructure Session) 5
Technical Program (Information and Communications Technology Session) 7
Technical Program (Manufacturing and Machinery Session) 8
Technical Program (Semiconductor, Materials and Electronics Session) 9
Book of Abstracts
Energy
Bjorn Ivan G. Ofrasio, Alexander F. Padilla
Jr., Tristan Jeorge Perez, Daryl Matthew
Energy Audit on Lighting and Air-conditioning
Tongson, Sophirak Nou, Karlo Jed Pestaño,
EGY_01 of Laboratories for Energy Cost Savings by 11
Jabar Mangandiri, Joseph Gerard T. Reyes,
Simulations in DIALux and eQuest
Carl Michael F. Odulio, Justin Jesse L.
Seranilla
Production and Characterization of Sol-Gel
Processed Microencapsulated Lauric Acid- Paolo Yves L. De Silos, Menandro S.
EGY_02 20
Based Phase Change Material from Treated Berana, Rizalinda L. De Leon
Waste Coconut Oil
Investigation of Heat Transfer Efficiency of
Tungsten Carbide and Cobalt Oxide Jhone Mychale I. Mariquit, Edgar Alan A.
EGY_03 24
Nanoparticles Dispersed on Distilled Water as Donasco
Base Fluid
Development of Hybrid Convection Solar Reniel Z. Rocaberte, Marvin M. Cinense,
EGY_04 31
Dryer for Seaweeds (Kappaphycus alvarezii S.) Ruel G. Peneyra, Jeffrey A. Lavarias
GIS-based Site Suitability Analysis for
Wenyville Nabor Galang, Ian Dominic
EGY_05 Biomass Energy Plant in Bohol's Rice and 36
Tabañag, Michael Loretero
Coconut Producer Municipalities
Environment and Infrastructure
Edgar Clyde R. Lopez, Jeanette Vivien R.
Colorimetric Detection of Copper Ions in Water Jalalon, Mary Faith C. Delas Alas, Cheska
EI_01 42
using Polymer-decorated Silver Nanoparticles Mae P. Aldana, Marlon L. Mopon, Jem
Valerie D. Perez
Geotechnical Characterization of Quarry Waste
EI_02 Limestone as an Alternative Road Base Course Emerzon S. Torres, Mary Ann Q. Adajar 44
Material
A Nonlinear Optimization Model for Subsidy
Allocation to Disrupted Infrastructures and Its
EI_03 Egberto F. Selerio, Jr. 46
Application in the Philippines during the
COVID-19 Pandemic
ERDT National Conference 2020
Theme:
CONCEPT NOTE
The pandemic has undoubtedly shaken our foundations. Our usual routines where we get to be
surrounded by peers and colleagues in the academe and in the workplace were replaced by new means of
conduct that are heavily anchored online.
As with precedent, science and technology continues to be more relevant in every passing day, but more
so during adversities. From the basic requirement of identifying the distance that should be observed by
people in order to prevent the transmission of the virus, to the larger demand for vaccine creation--- we
always rely in the field of science and technology for answers and solutions.
In this time of crisis, we are called upon to come back to the maxim, “Science and Technology for the
People,” to ensure that our outputs are relevant and easily transferable for the benefit of other members
of society.
We enjoin our researchers, scientists, and engineers to continue with their noble pursuit of helping the
country bounce back into the new normal, and recognize that their efforts will undoubtedly ripple into
waves that will enable us to create a better and more secure future.
1
TECHNICAL PROGRAM
Meeting ID : 843-5345-2720
Passcode : ENC2020EGY
Energy Session
07 December 2020
Session Chair : Dr. Luzvisminda M. Bellotindos
Moderator : Engr. Isabelo A. Rabuya
Panelists : Dr. Reynaldo C. Guerrero, Mr. Robert Alfie S. Peña,
Mr. Raymark C. Parocha, Dr. Marvin M. Cinense
JHONE MYCHALE I.
Investigation of Heat Transfer Efficiency of
MARIQUIT
10:55AM – 11:10AM Tungsten Carbide and Cobalt Oxide Nanoparticles
Mindanao State University-Iligan
Dispersed on Distilled Water as Base Fluid
Institute of Technology
4
TECHNICAL PROGRAM
Meeting ID : 817-9034-7704
Passcode : ENC2020EI
Environment and Infrastructure Session
08 December 2020
Session Chair : Dr. Sheila K. Ramos
Moderator : Dr. Bryan G. Alamani
Panelists : Dr. Mary Donnabelle L. Balela, Dr. Candy C. Mercado, Dr. Annaliza Rollon,
Dr. Maria Antonia Tanchuling
09:10AM – 09:20AM Log-in and Roll Call
09:20AM – 09:25AM Welcome Remarks and Introduction of Plenary Speaker
09:25AM – 09:55AM Plenary Talk: DR. ENRICO C. PARINGIT
09:55AM – 10:10AM Open Forum
10:10AM – 10:15AM Preparatory Break
5
Simulating the Impacts of Deficit Irrigation on
Corn Yield and Water Productivity using MARILYN S. PAINAGAN
11:45AM – 11:55AM
Coupled DSSAT CERES-Maize and Aquacrop University of the Philippines Los Baños
Models
11:55AM – 12:00NN Q&A
12:00NN – 12:55PM Lunch Break
Chi and Hypsometric Analysis for Watershed IMELIDA G. TORREFRANCA
12:55PM – 01:05PM
Prioritization and Conservation University of San Carlos
01:05PM – 01:10PM Q&A
Automated Furrow Irrigation System for
SHEILA G. GRIENGO
01:10PM – 01:20PM Maize (Zea mays L. var. saccharata)
Central Luzon State University
Production
01:20PM – 01:25PM Q&A
Physical and Mechanical Properties
MARY ANN N. AHALAJAL
Assessment of Lignocellulosic Rice Straw-
01:25PM – 01:35PM Mindanao State University-Iligan Institute
based Rigid Polyurethane Foam for Structural
of Technology
Sandwich Panel Cores
01:35PM – 01:40PM Q&A
DAVE JOSEPH E. ESTRADA
Silanated Coconut Oil-based 2K Waterborne
01:40PM – 01:50PM Mindanao State University-Iligan Institute
Polyurethane Coating for Corrosion Protection
of Technology
01:50PM – 01:55 PM Q&A
CIELO D. FRIANEZA
01:55PM – 02:05PM Qualitative Risk Assessment of Tailings Dam
De La Salle University
02:05PM – 02:10PM Q&A
02:10PM – 02:20PM Deliberation and Announcement of Best Paper
6
TECHNICAL PROGRAM
Meeting ID : 819-3345-4662
Passcode : ENC2020ICT
Information and Communications Technology Session
09 December 2020
Session Chair : Dr. Edwin Sybingco
Moderator : Engr. Jose Martin Maningo
Panelists : Dr. Kathleen Aviso, Dr. Argel Bandala, Engr. John Anthony Jose
08:45AM – 08:55AM Log-in and Roll Call
08:55AM – 09:00AM Welcome Remarks and Introduction of Plenary Speaker
09:00AM – 09:30AM Plenary Talk: DR. JOHN-JOHN CABIBIHAN
09:30AM – 09:45AM Open Forum
09:45AM – 09:50AM Preparatory Break
7
TECHNICAL PROGRAM
Meeting ID : 892-1144-3819
Passcode : ENC2020MM
8
TECHNICAL PROGRAM
Meeting ID : 875-5922-7833
Passcode : ENC2020SME
Semiconductor, Materials, and Electronics Session
11 December 2020
Session Chair : Dr. Jonathan W.L. Salvacion
Moderator : Engr. Ericson D. Dimaunahan
Panelists : Dr. Michael C. Pacis, Dr. Jennifer C. Dela Cruz,
Engr. Febus Reidj G. Cruz
09:10AM – 09:20AM Log-in and Roll Call
09:20AM – 09:25AM Welcome Remarks and Introduction of Plenary Speaker
09:25AM – 09:55AM Plenary Talk: DR. JOCELYN F. VILLAVERDE
09:55AM – 10:10AM Open Forum
10:10AM – 10:15AM Preparatory Break
PROF. FERDINAND G.
11:45AM – 12:00NN Closing Remarks MANEGDEG
Program Leader, ERDT
9
Energy
Session Chair : Dr. Luzvisminda M. Bellotindos
Moderator : Engr. Isabelo M. Rabuya
Panelists : Dr. Reynaldo C. Guerrero
Dr. Robert Alfie S. Peña
Mr. Raymark C. Parocha
Dr. Marvin M. Cinense
Energy audit on lighting and air-conditioning of
laboratories for energy cost savings by
simulations in DIALux and eQUEST
Bjorn Ivan G. Ofrasio1,a *, Alexander F. Padilla1,b, Jr., Tristan Jeorge Perez1,c, Daryl Matthew Tongson1,d, Sophirak
Nou1,e, Karlo Jed Pestaño1,f, Jabar Mangandiri1,g, Joseph Gerard T. Reyes1,2,h, Carl Michael F. Odulio1,3,i & Justin Jesse
L. Seranilla1,4,j
1
Energy Engineering Program, National Graduate School of Engineering, University of the Philippines Diliman
2
Department of Mechanical Engineering, University of the Philippines Diliman
3
Electronics and Electrical Engineering University of the Philippines Diliman
4
Institute of Civil Engineering, University of the Philippines Diliman
a
bjornivanofrasio09f@gmail.com bafpadilla@up.edu.ph ctj.perez23@yahoo.com.ph d darryl_tongson@hotmail.ph
e
nousophirak001@gmail.com f karlojedpestano@gmail.com g jabarmangandiri@gmail.com hjtreyes2@up.edu.ph
i
carl.odulio@eee.upd.edu.ph jjlseranilla@up.edu.ph
Abstract— Energy audit was conducted in two research Several problems in the UP Electrical and Electronics
facilities in the University of the Philippines Diliman, the Digital Institute (EEEI) laboratories, Microelectronics and Microprocessors
Signals Processing Laboratory (DSPLAB) and Microelectronics laboratory (MICROLAB) and the Digital Signal Processing
Laboratory (MICROLAB). DSPLAB was then divided into two Laboratory (DSPLAB), include dark spots, hot spots and thermal
energy activity centers (EAC). Inventory of products were discomfort to occupants overall. As such, energy audit was
conducted as a guide to the energy consumption of each laboratory. conducted on these rooms (1) to determine the causes of lighting
Illuminance and temperatures were sampled on grid-divided distributions on each room with the respective energy efficiency
laboratories for three different times (0900H, 1200H and 1500H) to index (EEI); (2) to determine the causes of temperature distributions
determine the lighting and temperature distributions of each on each room with the respective EEI; (3) to determine the cause of
laboratories at different times. It was found out that most lighting energy consumption profiles and inefficiencies on different times;
did not meet the 300-lux standard due to inefficiencies to lighting and (4) to recommend upgrading measures by computation and
caused by low rated lamp output, walls with low light reflectance, simulation based from the following criteria: lower EEI, monthly
layout of desks and layout of luminaires among others. energy savings and adhering to lighting and temperature standards.
Temperatures also did not pass the 23-25°C standard due to the
overrating or underrating of air-conditioning (AC) units compared
to the laboratory heat load. The layout of AC units also caused II. METHODOLOGY
certain areas to overcool and other areas uncooled. Likewise, energy
patterns suggest that energy consumption happens by noon to A. Materials
afternoon due to increase in number of occupants and sunlight A Bosch™ GLM 40 Laser Distance Meter was used to get
during morning. As visualized in DIALux™, recommended the room dimensions. A Smart Sensor digital lux meter AS403 with
measures for lighting include changing the wall paint, changing the a sampling rate of 1.5x per second and resolution of 1 lux was used
desk layout, rewiring the lights and replacing lamps into efficient to record the illuminance in lux. Benetech™ GM816A anemometer
ones that saves 312-1398 Wh/year and 1.3 to 7.2-year payback of was used to obtain both wind speed and temperature with 1-m from
investment. Recommended measures for air-conditioning was the air-conditioning unit. Creston™ FN-8830 clamp meter (double
simulated in eQuest™ which could save up to Php 197,363.00 and integral A/D switch implement measuring method) was used to
with maximum payback period of 5.1 years. measure electric current and voltage.
11
area. Air leakage sources were identified wherein holes or unsealed recommended minimum illuminance. 𝑁𝑅𝐿 is the number of
parts of the room were observed within the conditioned space remaining luminaires assuming the same lumens of the lamps.
boundary. Inventory of equipments (such as computers, laptops, DIALux™ lighting simulator software was used as a
battery charger, lamps and other electrical equipment) that simulation method for calculating LLF. The actual lighting
contributed to the heat load were recorded. Checking and listing the conditions were simulated to verify dark spots, to check the factors
power rating of each equipment provided an estimate on how high affecting the lighting and to propose measures to improve it.
or low is the typical power consumption of the facility. The Daylight effects is set to none to reduce the sunlight effects on the
condition of the equipment was checked for performance, leakage windows. This was also done for recommendations on the
for air-conditioning units, dimming for lamps and other issues. assumption that blinds and window tints will be applied for
The activity of the occupants for HVAC operation reduction of heat load.
schedule, lighting use and receptacle outlet use were also identified.
E. Heat load calculation and simulation
C. Energy audit and EEI measurement Heat loads are computed using (4)-(10) for solar heat
The energy audit was conducted from April 7 to May 16, gains (4), envelope heat gain (5) with resistance calculations if
2019. The audit schedule covered within operation days of Monday conductive (6) or convective (7), internal heat for luminaires (8),
to Friday only by which the measurements were taken three fixed internal heat for people (9) and internal heat for others (10) [2].
hours a day: 09:00 AM, 12:00 PM and 03:00 PM. Power Cooling load factor (CLF) for appliances is 1.0 since cooling system
consumption of the facility was obtained by taking current is not operated 24 hours. The equipment inventories (working
measurements and voltage measurements at the breaker room. Then, luminaires, idle and working computers, energized stand fans and
the lighting and receptacle loads of the facilities were measured on water dispenser) and occupancies of each laboratory were
the breaker room. Furthermore, the air conditioning loads are considered as internal heat sources. External heat sources were
measured beside their respective condenser units. Also, the current disaggregated into incident light and envelope conduction.
of the air conditioning units went from maximum rated current of Approximate values were used for the thermal resistance associated
17A to minimum operating current of 0.3A (operating blower). The to convection, these being for free air at a value of 2 𝑊𝑚−2 𝐾 −1 and
cycle time for the air conditioning units were identified to estimate static air at a value of 1 𝑊𝑚−2 𝐾 −1 . The average value of the EAC
its power consumption. Aside from the energy measurements, the was used as the temperature of the room. Recorded temperatures of
number of users of the laboratory during the audit period were the hallway and the bridge were used as a proxy for hallway and
counted to estimate the average occupancy and the heat load of the outside temperatures, respectively. A temperature of 24 ᵒC was used
room. for the adjacent rooms. Heat load was compared against the installed
Room temperatures of DSPLAB and MICRLAB were cooling capacity of the air conditioning unit to classify whether the
taken in the center of each grid at 0.76-m height (or table height) installed AC unit is overrated or underrated. A thermal circuit
using the anemometer. Other temperature measurements taken diagram was used to visualize the influences of heating load for its
include outside temperature and hallway temperature. adjacent areas.
The lux levels of the facility were taken simultaneously
with the temperature measurement. In light intensity measurement, 𝑄=𝐴 𝑥 𝑆𝐻𝐺𝐹 𝑥 𝑆𝐶 𝑥 𝐶𝐿𝐹 (4)
lux meter was placed at the center of each grid at the same level and
height as of the anemometer. 𝑄̇ = ∆𝑇/𝑅 (5)
EEI was obtained by plotting the raw data measurements
of temperature and luminance, tracing the best fit line and getting R=t/kA (6)
the value of power consumption at the reference illuminance level
(300 lux) or room temperature (24 ᵒC). The EEI from the raw data R=1/hA (7)
was used as the baseline for the recommendations.
𝑄=3.41* ( 𝑞𝑖 )*( 𝐹𝑢 )*( 𝐹𝑠 )*(𝐶𝐿𝐹) (8)
D. Lighting calculation and simulation
To verify the lighting measurements, Zonal Cavity 𝑄= n * [(SHG *CLF) + (LHG)] (9)
calculation was used to determine the variance between the
luminance of the installed lamp and the measured value [1]. 𝑄 = Heat Gain x CLF (10)
Parameters include room height, ceiling cavity height, room area,
room length and width and lamp rated output. (1)-(3) were used for 𝑄 is the heat load (W). A is the window area (m2). SHGF
calculating light loss factor (LLF) and total initial lamp lumens. is the solar heat gain factor. CLF is the cooling load factor. SC is the
shading coefficient. ∆𝑇 is the difference in temperatures (°C) in and
𝜙𝑡𝑜𝑡𝑎𝑙 × 𝐶𝑈 × 𝐿𝐿𝐹 out of the room. 𝑅 is the resistance. t is the wall thickness. k is the
𝐸𝐼𝑁𝑆𝑇 = (1)
𝐹. 𝐴. thermal conductivity of the wall. h is the convective heat transfer
coefficient. qi is the sensible heat gain from lights. Fu is the usage
𝐸 × 𝐴𝑟𝑒𝑎 factor. Fs is the service allowance factor. n is number of peopler.
𝑇𝐼𝐿𝐿 = (2)
𝐶𝑈 × 𝐿. 𝐿. 𝐹. SHG is the sensible heat gain of a person. LHG is the latent heat gain
of a person.
eQUEST™ was used to generate a more apt model. The
𝑇𝐼𝐿𝐿 room was considered in isolation. The floor plan was copied into the
𝑁𝑅𝐿 =
𝑛𝑜. 𝑜𝑓 𝑙𝑎𝑚𝑝𝑠 𝐼𝑛𝑖𝑡𝑖𝑎𝑙 (3) software. Wall construction was designated as 6-inch concrete,
[ ]×[ ]
𝑝𝑒𝑟 𝑙𝑢𝑚𝑖𝑛𝑎𝑖𝑟𝑒 𝑙𝑢𝑚𝑒𝑛𝑠 𝑝𝑒𝑟 𝑙𝑎𝑚𝑝 likewise with floor construction, ceiling construction and roof
construction. Floor insulation was set to be vinyl tiles as all EAC’s
𝐸𝐼𝑁𝑆𝑇 is the average illuminance on the surface-of- have vinyl tiles, whereas ceiling insulation was set to either be lay-
measurement. 𝜙𝑡𝑜𝑡𝑎𝑙 is the total lamp lumen output. 𝐶𝑈 is the in acoustic tiles, or none. Doors were set to be of wooden
coefficient of utilization. 𝐿𝐿𝐹 is the light loss factor. 𝐹. 𝐴. is the area construction with a hollow core, flushed at 1¾” and window
of the surface. 𝑇𝐼𝐿𝐿 is the total initial lamp lumens required. 𝐸 is the construction was set to single paned, clear at ⅛” thickness. A
weather data for Manila was provided by the software. Parameters
12
for occupancy and lighting was put into the software with lighting matches patterns of the color maps – indicating that the illumination
power consumption being expressed as watts per floor area. model for the MICROLAB is valid. The simulated lighting has a
However, it can neither model air flow nor specify layout of LLF of 0.51.
equipment and furniture within a room. As a result, it assumes that
the room is of even temperature all throughout and optimized air
flow.
F. Recommendations
The recommendations for energy efficiency were based
on the obtained EEI. The objective of the recommendation is to
maintain or lessen the energy consumption while increasing the
output of the energy utility appliance to its recommended level. The
simulations of lighting and air-conditioning were conducted using
DIALux™ and eQUEST™, respectively to determine the projected
consumption given a specific parameter. The floor plan, objects,
ceiling and walls were designed in the software. Recommendation
was classified as no cost, low cost or high cost depending on the
simple payback period as computed from (11) and (12).
𝑆𝑎𝑣𝑖𝑛𝑔𝑠
= (𝑘𝑊ℎ 𝑎𝑐𝑡𝑢𝑎𝑙 𝑎𝑛𝑛𝑢𝑎𝑙 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛
− 𝑘𝑊ℎ 𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑 𝑎𝑛𝑛𝑢𝑎𝑙 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛) (11)
𝑃ℎ𝑝
×( )
𝑘𝑊ℎ
𝐼𝑛𝑠𝑡𝑎𝑙𝑙𝑎𝑡𝑖𝑜𝑛 𝑐𝑜𝑠𝑡
𝑆𝑖𝑚𝑝𝑙𝑒 𝑝𝑎𝑦𝑏𝑎𝑐𝑘 𝑝𝑒𝑟𝑖𝑜𝑑 = (12) Fig. 1. Sample luminance data of MICROLAB (April 30, 2019) at (a) 9-10
𝑆𝑎𝑣𝑖𝑛𝑔𝑠
A.M., 115.8 lux average; (b) 12-1PM, 113.3 lux average; (c) 3-4PM, 117.6
lux average; and (d) average lighting simulation, 115 lux average.
III. RESULTS AND DISCUSSION
13
aging of the lamp which would affect the light loss factor resulting
to low lux output. This is apart from the occasional flickering of
some lamps.
DSPLAB EAC2 lighting was simulated with 12 lamps
with individual lamp rating of 2500 lumens (OSRAM 36W), 4
lamps with individual lamp rating of 2750 lumens (GE F36W T8),
2’ x 4’ x 3” PARAMAX parabolic troffer and a mounting height of
2.7 m with ceiling cavity of 0.65m. As shown in Fig. 3d, the dark
spots are mostly located on the portions of the room with busted
lamps showing lux levels reaching up to 25 lux as indicated on red
lines. Comparing with the values on the grid, the measured lux levels
are reaching 13 Lux on the portion of L10 and 60 Lux on the L8.
The simulation map matches patterns of the color maps – indicating
that the illumination model for the DSPLAB EAC2 is valid. The
LLF is at 0.49.
14
Table 1. Evaluation of existing air-conditioning equipment based on the cooling loads of each area.
None of the facilities reached the 24 °C target temperature. Table 2. Estimated energy consumption of each area.
This has caused thermal discomfort due to several factors. First,
doors between rooms are kept open as a space cooling practice of Description MICROLAB DSPLAB DSPLAB
the occupants. The sliding door between EAC1 and EAC2 was kept EAC 1 EAC 2
open with oscillating fans in between to keep the cold air circulating Average energy 5.004 1.993 0.861
between these areas. The doors leading the faculty are also kept consumption per
opened while the AC units are turned ON. hour (kWh)
Second, some of possible air leak sources include wall Average power in 39.3 15.9 6.9
conduits without proper sealing, edges of doors have no sealing and 8-h operating time
louvers leading to maintenance areas are not completely closed as (W)
well. Overtime instances 50 100 100
Third, placement of AC units are not efficient for even during audit (%)
cooling distribution of the space. In EAC1, the unit AC1 is placed at Hours of overtime 9 15 15
near the sliding door and at the end of the workstation which does (h)
not distribute air flow evenly within the workstation. In EAC2, the
Estimated daily 98.3 47.8 20.6
coldest spots are in the intersection of air flows of AC2 and AC3.
consumption
Most of the work desks are the coldest spots in the facility and
with overtime
reaching below 23 °C by afternoon. However, remaining areas did
(kWh)
not reach near 24 °C target (either lower than 23 °C or higher than
Cost per month* 15,654 10,839 4,680
25 °C) since placement of AC2 and AC3 and tables arrangement
(PhP)
made cooling uneven for the whole workspace. * 22 working days per month, 10 PhP/kWh.
Lastly, the unit AC3 does not have a definite temperature
as set point making temperature control difficult.
15
Table 3. Energy efficiency measures for lighting.
No. Facility Energy Conservation Measures Total Average EEI Initial Annual Payback
Installed Illuminance (W- Cost Energy Period
Rated (lux) hr/300 (Php) Savings (years)
Power lux) (W-hr/year)
(W)
EEM- MICROLAB Rearrange tables; lowered luminaire 300 348 300 23,092.80 312 7.2
L1 mounting to 2.3 m; wall paint
application; rewiring connection of
switch of L3; replace lamps with 15W
led 2000 lm
EEM- DSPLAB Add 2 luminaires; replace all lamp of 180 335 180 18,754.00 449.28 4
L2 EAC1 luminaires to 18 W, 2100 lm; rearrange
mounting of lamps; apply white wall
paint; installed ceiling panels
EEM- DSPLAB Repaired 4 busted lamps; applied white 352 320 352 16,587.00 1,397.76 1.3
L3 EAC2 wall paint; replace lamps with LED
2640-lm 22-W; installed ceiling panels
It was observed that the three areas have similar energy luminaires and applying white wall paint. This has led to a lower
consumption characteristics. MICROLAB, DSPLAB EAC1 and power consumption and uniform distribution of 300 lux on the
EAC2 have 4.48, 1.55 and 0.63 kWh morning consumptions, workstation. This will lower EEI to 180 Wh/300 lux from 216
respectively. Morning (0900H-1000H) has the lowest energy Wh/300 lux. It requires an investment of PhP 18,564 but the energy
consumption. This was due to the cool temperature felt in the savings per month will be at 449.28 Wh.
morning which delays AC starting operation, closed lamps if the Lighting recommendation for DSPLAB EAC2 (EEM-L3)
EAC is hit by sunlight, low morning activity of the occupants and is illustrated in Fig. 5c. This alternative includes replacing all lamps
low occupancy overall. to WY™ 4FT T8 LED Tube (22-W, 2640-lumen LED), painting of
Noontime (1200H-1300H) and afternoon (1500H-1600H) walls into white and installing ceiling panels. This will cause an
energy consumptions for MICROLAB (5.35 to 5.20 kWh), average luminance of 320 lux and lower EEI to 352 Wh/300 lux.
DSPLAB EAC1 (2.21 to 2.25 kWh) and EAC2 (1.29 to 1.42) are This is a better alternative than a low-cost measure to replace the
almost similar to each other but are larger than the morning energy busted lamps L8 and L10 but will increase EEI to 648 Wh/300 lux
consumption. This is due to the maximum occupancy of these areas aside from the 192-lux average illuminance. The cost of upgrade is
resulting to maximum power output from the computers, full AC Php 16,588 with 1.3-years payback and 1398-Wh monthly energy
and lighting operations. savings.
Accounting for all energized equipment (lighting,
receptacle and AC), the estimated energy consumption within the E. Air-conditioning recommendations
audit period was determined. Basted from Table 1, MICROLAB has Average data for MICROLAB (115.6 lux, 25.9 °C) was
the most energy consumption with 5 kWh. This is due to the desktop used in eQUEST™ to get the annual energy consumption (16,309.09
computers operating for 24 hours on active and idle modes. kWh). Savings were based on differences in power consumption
However, daily consumption increases due to overtime activities of from the simulations applying the energy conservation measures.
the occupants doing research and study activities. This has led to Energy consumption is assumed to have a rate of Php 10/kWh.
daily monthly electricity costs to around PhP 15654, PhP 10839 and EEM-T1. All energy conservation measures are
PhP 4680 for MICROLAB, DSPLAB EAC1 and DSPLAB EAC2, summarized in Table 4. Tint and roll-up blinds were applied to
respectively. windows to reduce heat fenestration. Door, window and conduit
seals were applied to solve the problems on air leakage. Simulation
D. Lighting recommendations of air-flow optimization will have a projected consumption of
Table 3 lists the recommended measures of the three areas. 13,590.91 kWh per year. The air-flow optimization alone will cost
A high cost alternative is proposed for the MICROLAB (EEM-L1) Php 43,053.60, will lower EEI to 2613.63 W-h/24 °C and will have
as illustrated in Fig. 5a. This recommendation involves low-cost a 0.6-year payback.
methods: rearranging desks and lowering of light mount; and high- EEM-T2. A high cost solution involves replacing of AC2
cost methods: replacement of lamps to 15-W 2000 lm (LIFE and AC3 by a 2-TR Carrier™ ceiling cassette AC unit (FP-
BULB™ 4-ft. T8 LED Tube), painting all walls with white latex to 53CCFASB024308-2T). This will give appropriate cooling load to
increase LRV, and rewire lamp L3 to the workstation lamp circuit. the MICROLAB and give proper air flow to working areas as shown
This will lower EEI by 60 Wh into 300Wh/300 lux with a uniform in Fig. 6a. Simulation of AC installation will have a projected
distribution of 300 lux on the workstation. The lighting simulation consumption of 8,981.82 kWh per year. Air-flow optimization with
results are shown on Fig. 5. installation of cassette-type ACU will cost Php 27,181.82, will lower
A recommendation for the DSPLAB EAC1 is a high-cost EEI to 1727.27 W-h/24 °C and will have a 5.1-years payback.
alternative (EEM-L2) illustrated in Fig. 5b. Low cost measures 3.5.2 DSPLAB EAC1
include rearrangement of luminaires by 90°. Adding luminaires and
rearranging them is inevitable since the tables are placed on the
edges of the room. Rearrangement of computers is not applicable as
well as the room width is cramped. High cost measures include
replacing lamps to New Lights™ 18-W, 2100-lumen LED, adding 2
16
Table 4. Energy efficiency measures for thermal control and air flow.
No. Facility Energy Conservation Measures Annual EEI Initial Cost Annual Payback
Consumption (W- (PHP) Savings Period
(kW-hr/year) hr/24ᵒC) (PHP) (years)
EEM- MICROLAB Airflow Optimization (Seals) 13,590.91 2613.63 43,053.60 73,272.73 0.6
T1 Addition of Tints and Blinds
EEM- AC3 Replacement (2TR cassette AC) 8,981.82 1727.27 139,892.40 27,181.82 5.1
T2 Airflow Optimization (Seals)
Addition of Tints and Blinds
EEM- DSPLAB Airflow Optimization (Seal) 4,466.09 715.72 27,420.00 29,675.45 0.9
T3 EAC1 AC1 Relocation
EEM- Ceiling Insulation 561.73 710.417 39,084.00 30,006.36 1.3
T4 Airflow Optimization (Seals)
Addition of Blinds
AC1 Relocation
EEM- Lowered Ceiling Insulation 4,801.73 721.78 40,812.00 29,687.00 1.4
T5 Airflow Optimization (Seals)
AC1 Relocation
EEM- DSPLAB AC Repair 12,290.91 1969.697 37,314.00 185,545.50 0.2
T6 EAC2 Airflow Optimization (Seals)
Addition of Tints and Blinds
EEM- Lowered Ceiling 12,172.73 1950.758 68,850.00 186,727.30 0.4
T7 AC Repair
Airflow Optimization (Seals)
Addition of Tints and Blinds
EEM- AC3 Replacement (floor-mounted) 11,109.09 1780.303 146,848.80 192,363.60 0.8
T8 Lowered Ceiling
Airflow Optimization (Seals)
Addition of Tints and Blinds
EEM- AC3 Replacement (3TR cassette AC) 11,109.09 1780.303 208,768.80 197,363.60 1.1
T9 Lowered Ceiling
Airflow Optimization (Seals)
Addition of Tints and Blinds
EEM- AC3 Replacement (two 1.5TR cassette 11,109.09 1780.303 188,550.00 197,363.60 1.0
T10 AC)
Lowered Ceiling Insulation
Airflow Optimization (Seals)
Addition of Tints and Blinds
Using actual average data of 175.5 lux lighting and 22.1 721.78 W-h/24 °C. This will cost PHP 40,812.00 and will be paid
°C temperature, the power consumption simulated in eQUEST™ is back in 1.4 years.
7,433.64 kWh per year. Energy consumption rate is assumed to be For DSPLAB (122.5 lux, 24.4 °C), the annual energy
Php10/kWh. Savings were based from differences in power consumption simulated is 30,845.45 kWh in a year. The layouts in
consumption from the simulations applying the energy conservation Fig. 6b, c and d are assumed to have an optimized air flow.
measures. Air flow optimization is assumed by following the EEM-T6. Tint and roll-up blinds were applied to windows
proposed layout in Fig. 6c and d. to reduce heat fenestration. Door, window and conduit seals were
EEM-T3. One-meter roll-up blinds were applied to applied to solve the problems on air leakage. AC3 repair will also
windows without tint to reduce heat fenestration. Door, window and optimize air flow. Simulation of air-flow optimization alone will
conduit seals were applied to solve the problems on air leakage. have a projected consumption of 12,290.91 kWh per year. The air-
Simulation of air-flow optimization alone will have a projected flow optimization alone will cost Php 37,314.00, will lower EEI to
consumption of 4,466.09 kWh per year. The air-flow optimization 1969.697 W-h/24 °C and will have a 0.2-of-a-year payback. Layout
alone will cost PHP 27,420.00, will lower EEI to 715.72 W-h/24 °C shown in Fig. 6b.
and will have a 0.9-of-a-year payback. EEM-T7. Aside from AC3 repair, addition of ceiling
EEM-T4. Installing ceiling insulation will give better insulation 2.3-m high will give better airflow around DSPLAB
airflow around DSPLAB EAC1 and insulating the room further EAC1 and insulating the room further from external heat source.
from external heat source. Simulation of AC installation will have a Simulation of AC installation will have a projected consumption of
projected consumption of 39,084.00 kWh per year. Air-flow 12,172.73 kWh per year. Air-flow optimization with ceiling
optimization with ceiling insulation will cost PHP 30,006.36, will insulation will cost Php 68,850.00, will lower EEI to 1950.758 W-
lower EEI to 710.417 W-h/24 °C and will be paid back in 1.3 years. h/24 °C and will be paid back in 0.4-of-a-year. Layout shown in Fig.
EEM-T5. Applying ceiling insulation 2.7-m above with 6b.
conduit seals, window blinds and tints will give DSPLAB EAC1 a EEM-T8. Replacing of AC3 with a 3TR QUBE™ floor-
projected consumption of 40,812.00 kWh per year. EEI will be mount split-type (SKU11415) AC unit with ceiling installation will
17
Fig. 6. Proposed layout on (a) MICROLAB; (b) DSPLAB with AC3 repair or replacement, (c) DSPLAB with cassette-type AC installation, and (d)
DSPLAB with two ceiling-mounted installation.
have a projected consumption of 11,109.09 kWh per year. It will overall layout of the office desks. Temperature distribution
reduce the EEI to 1780.303 W-h/24 °C by spending Php 146,848.80 inefficiencies are found due to the AC layout, overrating or
in total and will be paid back in 0.8-of-a-year. Layout shown in Fig. underrating of the installed AC units and service length of AC units.
6b. This has caused inefficiencies in power use and, therefore, increase
EEM-T9-10. Replacing of AC3 with a 3-TR Carrier™ in monthly electricity costs. Energy consumptions are the lowest in
cassette-type AC unit (FP-53CAV036308) (EEM-T9) or two 1.5- daytime and highest during noon to afternoon. This activity extends
TR LG™ cassette-type AC unit (JTNQ18GPLE5) (EEM-T10) will when occupants file overtime occurring after 1700H.
have a projected consumption of 11,109.09 kWh per year same as Energy efficiency measures (EEM) were recommended
replacing with floor-mount AC. This will reduce the EEI to for lighting and temperature. These measures adhere to the
1780.303 W-h/24 °C by spending Php 208,768.80 in total and will following criteria: lower EEI, monthly energy savings and outputs
be paid back in 1 to 1.1 years. Layouts show in Fig. 6c and d. with at least 300 lux and within 23-25 °C. The keys for an energy-
efficient light distribution include the following: desk alignment to
IV. CONCLUSION luminaires, high reflective wall color, installation of ceiling panels,
and replacement of lamps to a lower power rating but can supply
In conclusion, energy audit was done to recommend 300 lux at least. An energy-efficient temperature distribution
energy saving upgrades visualized by their costs. Lighting includes the following upgrades: sealing air leaks, window tints and
distributions is generally inefficient because of low lamp lighting blinds, ceiling insulation, layout of AC unit, correct AC unit
output, unreplaced busted lamps, low reflectance of wall colors and
18
installation based on the heat load of the room, and (optional) use of
a cassette-type AC unit for a more even cooling distribution.
ACKNOWLEDGEMENT
REFERENCES
19
Production and Characterization of Sol-Gel Processed
Microencapsulated Lauric Acid-Based Phase Change
Material from Treated Waste Coconut Oil
Paolo Yves L. De Silos1, a *, Menandro S. Berana,b, and Rizalinda L. De Leon3,c
1
Energy Engineering Program, College of Engineering,
University of the Philippines Diliman, Quezon City, Philippines
2
Department of Mechanical Engineering, College of Engineering,
University of the Philippines Diliman, Quezon City, Philippines
3
Department of Chemical Engineering, College of Engineering,
University of the Philippines Diliman, Quezon City,
a
pldesilos1@up.edu.ph, bmenandro.berana@coe.upd.edu.ph, crizalinda.deleon@coe.upd.edu.ph
* please mark the corresponding author with an asterisk
Abstract— The study dealt with the production and fluctuations of temperature and reduce the load of the cooling
characterization of microencapsulated lauric acid from treated waste equipment for energy efficiency [1].
coconut oil. The treatment of the waste coconut oil included Coconut oil is widely used in the Philippines as a primary
filtration, water removal, acid-catalyzed esterification, and alkali- cooking material. Because of the abundance in food processing
catalyzed transesterification. For the extraction of lauric acid, the plants in the country, a large amount of waste cooking oil is
method of Ezigbo & Mbaegbu which used sodium hydroxide was produced in a day-to-day period. Likewise, there is a need to
applied. The extracted lauric acid was encapsulated using Sol-Gel determine possible uses of this waste product to promote sustainable
process. In the preparation of the mPCM, the effect of drying time and interconnected waste management and renewable source of
and drying temperature on the product weight, diameter, energy. The formulation and production of cost-effective,
encapsulation ratio, and latent heat of melting was determined. For environmental-friendly, and sustainable ways in creating PCM is
the characterization, thermal properties were determined using vital to ensure waste minimization and energy security. There is a
Differential Scanning Calorimetry (DSC), presence of possible compelling reason to use waste cooking oil from food processing
functional groups were identified using Fourier Transform Infrared plants in order to extract lauric acid which can be microencapsulated
(FTIR) Spectroscopy, and the morphology were investigated using to form thermal energy systems [2][3].
Scanning Electron Microscopy (SEM). In FTIR, a greater detection In general, this study aims to produced sol-gel processed
of lauric acid may be inferred from the higher transmittance. microencapsulated lauric acid-based phase change material from
Likewise, a lower detection of lauric acid in the microencapsulated treated waste coconut oil and determine its properties as thermal
PCM may be attributed to the ineffective encapsulation which may energy system (TES). Specifically, this study aims to: determine the
result to leakage. The peaks for the functional groups of lauric acid effect of drying temperature and drying time in the product weight,
and silicate were found. On the other hand, a faster drying with high diameter, encapsulation ratio, and latent heat of melting of the
temperature result to a high yield. Among the lauric acid sources, microencapsulated phase change material (mPCM) from the
the one from technical grade lauric acid obtained the highest due to extracted lauric acid obtained from the treated waste coconut oil;
purity. In the DSC, a fast drying with high temperature resulted to a characterize the morphology of the mPCM using Scanning Electron
low enthalpy. Likewise, at constant drying time, increasing the Microscopy (SEM); identify the functional groups present in the
drying temperature resulted to an increased in latent heat of melting mPCM using Fourier Transform Infrared (FTIR) Spectroscopy; and
and a less steep melting temperature. In addition, in using SEM it determine the thermal properties of the mPCM using Differential
was seen that mPCM produced from low temperature and slower Scanning Calorimetry (DSC).
drying condition resulted to longest diameter due to longer time for
agglomeration. The significance of the drying time and drying
temperature in the product weight, diameter, and encapsulation ratio II. METHODOLOGY
was studied. It was seen that the drying time is significant in the
product weight and diameter. However, no conclusion was made on A. Materials and Methods
its significance in the encapsulation ratio and latent heat of melting. The methods included in the study were the treatment of
Moreover, the drying temperature was seen as significant in all four the waste coconut oil, the extraction of lauric acid, the
parameters. microencapsulation of lauric acid, and the characterizations of the
Keywords—lauric acid, phase change material, coconut mPCM. The reagents used were obtained from Sigma Aldrich and
oil, sol-gel, waste-to-energy the analyses were done at the Department of Science and
Technology – Advanced Device and Materials Testing Laboratory
(DOST-ADMATEL) and the UP Diliman Department of Chemical
I. INTRODUCTION Engineering. For the treatment of waste coconut oil, filtration, water
removal, acid-catalyzed esterification, and alkali-catalyzed
Various studies have already been done to determine the transesterification were done [4]. For the acid-catalyzed
capability of forming phase change material (PCM) from lauric acid. esterification, 2 mL 95% sulfuric acid, 6:1 methanol-to-oil molar
Variations include the combination with other fatty acids or ratio, 120-min reaction time, and 60 °C reaction temperature were
catalysts. The unique characteristics of lauric acid that makes it applied. Moreover, for the alkali-catalyzed transesterification, a 5.5
viable for further studies are its low vapor pressure, low degree of g/L KOH-to-oil ratio, 5:1 methanol-to-oil molar ratio, 60-min
supercooling, high heat of fusion, and compatibility with reaction time, and 60 °C reaction temperature were used. The
encapsulating material. Hence, it can prevent, or at least delay, the
20
forming glycerol was removed using a separatory funnel to avoid E WCO 12 75
unwanted products in the succeeding reactions [5][6]. F TGCO 12 75
For the extraction of lauric acid, a 10 M solution of NaOH G TGLA 12 75
was prepared. In a heated water bath, the treated coconut oil and
NaOH were mixed. Afterwards, the coconut oil was converted into
glycerine and lauric acid and a distinct two-layer immiscible mixture
was observed. The lauric acid was separated using a separatory
III. RESULTS AND DISCUSSION
funnel. The glycerine by-product was removed from time-to-time.
To confirm the presence of lauric acid, a two-g sample was placed A. Treatment of Waste Coconut Oil
in a beaker and acidified KMnO4 was added. The lauric acid
decolorized the acidified KMnO4 completely; thus, confirming the Pretreatment of the waste coconut oil was needed to
presence of lauric acid [7]. prevent the formation of unwanted by-products due to the reduction
The microencapsulation method used was Sol-Gel of transesterification productivity caused by the impaired oil quality
Process. In a beaker containing 50 mL distilled water, a 43 mL [12]. The importance of transesterification in the production of fatty
tetraethyl orthosilicate (TEOS) and 50 mL absolute ethanol were acid methyl ester (FAME) involves the reaction between short-chain
mixed. The beaker was placed in a hot plate with magnetic stirrer. alcohol (e.g. methanol or ethanol) and triglycerides (oil), using an
The alkoxide solution was agitated at 300 rpm. A 30-mL distilled appropriate catalyst, to produce fatty acid alkyl esters (biodiesel) and
water and sulfuric acid solution was prepared on a separate beaker glycerine. Transesterifications may be differentiated in terms of the
before being mixed with the alkoxide solution. To ensure catalyst used as either acid-catalyzed, alkali-catalyzed, or enzyme-
miscibility, the mixture was heated to 65 °C for 90 min and agitation catalyzed. Enzyme-catalyzed transesterification is impractical due
speed was increased to 600 rpm to prevent agglomeration. The sol to its slow reaction and high cost. The acid-catalyzed esterification,
mixture was separated after the completion of the hydrolysis and which used sulfuric acid as catalyst, was needed since WCO has a
condensation reactions of tetraethyl silicate. A 24-g lauric acid was high composition of FFA. The additional acid treatment converted
added to the sol and stirring at 800 rpm for 40 min was needed to the FFA to ester [4][6].
ensure uniform dispersion within the mixture. The mixture
temperature was maintained at 80 °C. The sol was poured into a B. Physical Properties of mPCM
slightly uncovered glass Petri dish and was placed in a gravity
Removal of water prior to the microencapsulation process
convection oven using a combination of drying temperature and
ensures a higher product weight. Likewise, a lower yield is obtained
drying time [8][9][10]. After the drying process, the glass Petri dish
because the high water component in the mixture increases the sol-
was removed from the oven and stored in a desiccator. Gel time, or
gel condensation. At low pH level, hydrolysis occurs slowly. Linear
the storage time of the glass Petri dish in the desiccator, was 5 days
molecules that are cross-linked are formed from the silica.
since the acid catalyst used was sulfuric acid. The product was the
Moreover, the entanglement of the molecular chains result to the
PCM which was microencapsulated.
additional branching [8]. Consequently, gelation occurs. Of the
samples which source is the treated waste coconut oil, the one with
B. Characterization of mPCM fast drying and high temperature got the highest yield. On the other
hand, comparing the samples with the same drying conditions but
The In determining the effect of drying temperature and different sources, the one from the technical grade lauric acid had
drying time in the characteristics of the microencapsulated lauric the highest weight. This can be attributed to the purity and
acid-based PCM, 7 runs were designed. Sample A did not undergo abundance of the lauric acid due to its grade.
drying, while Samples B, C, D, E, F, and G underwent drying with The morphology of the mPCM was studied using a
varying drying time, drying temperature combination, and lauric Hitachi S-3400 Scanning Electron Microscope. The SEM images or
acid source. The mPCMs were formed using either the waste micrographs of the samples were exposed in a 15 kV electron beam
coconut oil (WCO), technical grade coconut oil (TGCO), or with a magnification of x50, x500, x1000, and x5000. The
technical grade lauric acid (TGLA). For the characterization, DSC, agglomeration of non-homogenous microparticles or ‘multicore
FTIR Spectroscopy, and SEM were applied. The response variable structure’. The microencapsulated lauric acid can be assumed by
that was monitored in each run was the encapsulation ratio, product searching for the lumping mass in the middle of each particle. From
weight, latent heat of melting, and diameter. The encapsulation ratio the mPCM products, the one with low temperature and longer drying
is the ratio of the latent heat of meting of lauric acid and the time, had the highest diameter. This may be attributed to the longer
microencapsulated PCM [8][11]. agglomeration period. On the other hand, the samples with faster
drying time and medium to high drying temperatures got a lower
∆𝐻𝑚,𝑚𝑃𝐶𝑀
𝑅= 𝑥 100 (1) diameter.
∆𝐻𝑚,𝐿𝐴
where:
R = encapsulation ratio (%)
ΔHm,LA = latent heat of melting of the lauric acid (J/g)
ΔHm,mPCM = latent heat of melting of the mPCM (J/g)
21
C. Chemical Properties of mPCM A fast drying with high temperature resulted to a low
latent heat of melting. Hence, drying using the 7 hours drying time
The chemical compounds present in the mPCM was is more practical. Moreover, the melting temperature of the product
investigated by applying FTIR. This was needed in confirming the was affected by the drying temperature. For the latent heat of
encapsulation of the lauric acid core material by the silica-based melting of the mPCM, at constant drying temperature, an increased
shells. The spectroscopy was conducted from 0 to 4000 cm-1 in the drying time resulted in an increased in the latent heat of
wavenumber. From the IR spectra, there is a consistency in the dips melting of the mPCM; while at constant drying time, an increased
found from 0 to 2500 cm-1 from the various microencapsulated in the drying temperature also resulted in an increased in the latent
samples. However, beyond 2500 cm-1, deviations were observed. heat of melting of the mPCM.
This may be attributed to the varying concentration of lauric acid in
the samples. Transmittance is the ratio of the light energy entering E. Statistical Analysis
on a body to that transmitted through the body. In the IR spectra, a
greater detection of lauric acid may be inferred from the higher The significance of the drying time and drying
transmittance. Likewise, a lower detection of lauric acid in the temperature in product weight, diameter, encapsulation ratio, and
microencapsulated PCM may be attributed to the ineffective latent heat of melting of the microencapsulated PCM was analyzed
encapsulation which may result to leakage. The tetraethyl using Two-Sample T-Test in Minitab. A 95% confidence interval
orthosilicate peaks are at 463 cm-1, 960 cm-1, and 1083 cm-1. The was applied; hence a calculated p < 0.05 infers a significance.
hydrolysis-polycondensation reactions were catalyzed by acid Applying the analysis, it was found that the drying time of the
nucleophilic substitution reactions. There is no chemical interaction mPCM is significant for the product weight and diameter. Likewise,
between the silicate and lauric acid. Likewise, surface tension of the the drying temperature is significant for the product weight,
silica-based shell was retained which prevents leakage of the lauric diameter, encapsulation ratio, and latent heat of melting of the
acid core material. Moreover, silanol (Si-OH) and siloxane (Si-O- mPCM. Additional runs and specific statistical tools are needed to
Si), both basic groups, are responsible for hydrogen bonding which determine if drying time is not significant in the encapsulation ratio
produces a negatively-charged surface. and latent heat of melting of the mPCM.
22
addition, in using SEM it was seen that mPCM produced from low REFERENCES
temperature and slower drying condition resulted to longest
diameter due to longer time for agglomeration. [1] Alva, G., Lin, Y., Liu, L., & Fang, G. (2017). Synthesis,
Using T-Test, it was seen that the drying temperature has characterization and applications of microencapsulated phase
significant effect on product weight, diameter, encapsulation ratio, change materials in thermal energy storage: a review. Energy
and latent heat of melting of the mPCM. Moreover, it was seen that and Buildings, 144, 276-294
drying time is significant on the product weight and diameter. [2] Dimaano, M. N. R., & Watanabe, T. (2009). The capric-lauric
However, no conclusion was made for its effect on the encapsulation acid and pentadecane combination as phase change material for
ratio and latent heat of melting of the mPCM. low-temperature thermal energy storage
An optimization study on the effect of drying time and [3] Pielichowska, K., & Pielichowski, K. (2014). Phase change
drying temperature on the production of microencapsulated PCM materials for thermal energy storage. Progress in materials
from waste coconut oil is recommended. Moreover, a change in the science, 65, 67-123
waste feedstock is suggested due to its similar abundance in Metro [4] Duti, I. J., Maliha, M., & Ahmed, S. (2016). Biodiesel production
Manila and nearby provinces. Likewise, the effect of varying pH from waste frying oil and its process simulation. J. Mod. Sci.
level, water content, and oil ratio may also be looked in to. Technol, 4(1), 50-62
[5] Ullah, Z., Bustam, M. A., & Man, Z. (2014). Characterization of
waste palm cooking oil for biodiesel production. International
ACKNOWLEDGEMENT Journal of Chemical Engineering and Applications, 5(2), 134
[6] Adepoju, T. F. & Olawale, O. (2014). Acid-catalyzed
Utmost appreciation is given to the Department of Science esterification of waste cooking oil with high FFA for biodiesel
and Technology – Engineering Research and Development for production. Chemical and Process Engineering Research, 21,
Technology (DOST-ERDT) for funding the research study. 80-85.
[7] Ezigbo, V. & Mbaegbu, E. (2018). Extraction of Lauric Acid
from Coconut Oil, Its Applications and Health Implications on
Some Microorganisms. African Journal of Education, Science
and Technology, 3(3), PP 59-62
[8] Gecain, M. K. R. (2015). Preparation and properties of
microencapsulated lauric acid phase change material
incorporated in glass sheet for thermal regulation (Unpublished
master’s thesis). University of the Philippines - Diliman, Quezon
City, Philippines
[9] Sciortino, M., Alonzo, G., Ciriminna, R., & Pagliaro, M. (2011).
Sol–Gel microcapsulation in silica-based particles: A
comparative study. Silicon, 3(2), 77-83
[10] Brodhacker, K. L. (2011). The sol-gel process [Powerpoint
slides]. Lander University, South Carolina, USA
[11] Fang, G., Li, H., Chen, Z., & Liu, X. (2011). Preparation and
properties of palmitic acid/SiO2 composites with flame retardant
as thermal energy storage materials. Solar Energy Materials and
Solar Cells, 95(7), 1875-1881.
[12] Guerrero-Romero, A., & Sierra, F. E. (2011). Biodiesel
production from waste cooking oil. In Biodiesel-Feedstocks and
Processing Technologies. IntechOpen.
23
Investigation of Heat Transfer Efficiency of Tungsten
Carbide and Cobalt Oxide Nanoparticles Dispersed on
Distilled Water as Base Fluid
Jhone Mychale I. Mariquit1, a *, Edgar Alan A. Donasco2, b *
1
Luinab, Iligan City, Lanao del Norte 9200, Philippines
2
Xavier Heights, Cagayan de Oro City, Misamis Oriental 9000, Philippines
a
jhonemychale.mariquit@g.msuiit.edu.ph, bedgaralan.donasco.g.msuiit.edu.ph,
Abstract—Dealing with the inquiries on the effects on nanoparticles were used in the study [3] to improve the thermal
heat transfer efficiency of nanofluids with water as base fluid at conductivity of fluids. Single type nanoparticle suspensions
laminar and transient regimes, Tungsten Carbide (WC), and Cobalt displayed the highest improvement in thermal conductivity at
(II, III) Oxide (Co3O4), were tested in a shell and tube type crossflow around 74% increase compared to that of the base fluid. Between
heat exchanger with 0.25-inch inner diameter. Percent volume nanoparticles considered, copper nanoparticle showed the greatest
concentrations of nanoparticles ranges from 0.5% to 2.0% in improvement. An environment - friendly process utilizing intensive
increments of 0.5%. Impacts on the thermo – physical properties solar electromagnetic radiation is used to synthesize graphene and
were investigated. Considered also was the comparison of Nusselt graphene–multi walled nanotube (MWNT) hybrid nanoparticle [4].
number between theoretical, experimental, and computer simulated For 0.04% volume concentration, thermal conductivity
results. The study showed that adding nanoparticles had substantial improvement of 9.2% with graphene nanofluids was observed and
impacts. Decrease in heat capacity can be perceived for all 10.5% increase for graphene–MWNT nanofluids dispersed in de-
nanofluids investigated. While increase in thermal conductivity, ionized water considered at room temperature. Additionally, at
viscosity, and density was recorded. Theoretical computations Reynolds number of 2000, an improvement in heat-transfer
compared with computer simulated results had the highest coefficient of 193% was observed for 0.02% volume fraction of
discrepancy. While experimental data and simulated results had graphene–MWNT hybrid nanofluids. A study [5] which defines an
smaller difference. Regarding the effects on heat transfer efficiency, innovative synthesis method for a hybrid nanostructure comprising
WC nanofluid presented deteriorations in all particle loading even of multi walled nanotubes, hydrogen exfoliated graphene and silver
though it exhibited boost in thermal conductivity. While cobalt nanoparticles measured the thermal conductivity and convective
oxide nanofluid displayed substantial improvements, with the heat transfer characteristics at varied volume concentrations. An
highest augmentation recorded at 13.27% for 1.5% volume improvement of around 8% in thermal conductivity is attained for a
concentration. Based on the experimental data, correlations are volume concentration of 0.04% while improvement in heat transfer
proposed in approximating the Nu of the nanofluids investigated coefficient of nearly 570% for 0.005% volume fraction was
with ±10% consistency between predicted and experimental values. reported. Friction factor as well as convective heat transfer
characteristics of Al2O3–water nanofluid in a round tube with
Keywords—heat transfer, laminar flow, nanofluids, uniform wall temperature on turbulent flow settings were considered
transient condition in an experiment [6]. An average of 25% rise in heat transfer
coefficient was detected but with a 20% drawback in pressure drop.
I. INTRODUCTION As can be perceived, the intensification in heat transfer coefficient
in nanofluids paralleled to base fluids can vary from a few percent
The development of convective heat transfers and the up to several hundreds of percent. Nonetheless, there are studies
associated investigational as well as experimental and theoretical where the heat transfer efficiency decreases or there is no significant
research turn out to be at present a self-regulating, significant and improvement observed with the supplement of nanoparticles to the
rapidly developing field of heat transfer theory. base fluids.
High heat flux elimination or removal is a key attention in An experimental research [7] considered alumina and
the design of many machineries, apparatus and technologies, and zirconia nanoparticles. Alumina nanoparticle concentration was
can be achieved using several types of heat transfer equipment. High varied from 0.9 – 3.6% while zirconia nanoparticle was varied from
heat-flux cooling methods have obtained great consideration from 0.2 – 0.9%. The test revealed no unusual heat transfer improvement.
researchers due to its standing in thermal management of both A different set – up [8] which considers natural convection of
commercial and defense high-power electronic devices. While nanofluids inside horizontal cylinder reported a seemingly
remarkable development has been made in the past, high heat-flux contradictory behavior of heat transfer deterioration. Nanoparticles
removal still principally remains as a stimulating topic that needs utilized in the study were aluminum oxide and copper oxide with
additional investigation and study [1]. One of the several answers to variation of percent volume concentration of 1% and 4%. According
the problem of heat transfer performance improvement could be the to the report, the decline was dependent on particle density,
use of the so called nanofluids, which are basically fluids containing concentration and the aspect ratio of the cylinder. Titanium oxide –
nanoparticles of countless composition. Nanofluids can be deemed water nanofluid with particle sizes ranging from 30 – 40 nm with
as the new heat transfer fluids for they propose stimulating potentials volumetric concentration of up to 3% was investigated
to improve heat transfer performance against pure liquids. They are experimentally [9]. The findings also revealed that the
expected to have superior properties compared to conventional heat supplementation or presence of nanoparticles in the base fluid
transfer fluids [2]. Carbon nanotubes, copper nanoparticles and gold decreases the natural convective heat transfer coefficient within the
nanoparticles, along with hybrid nanoparticles such as carbon conditions of the investigation. Considering hybrid nanoparticles, an
nanotube–copper nanoparticles or carbon nanotube–gold experimental investigation [10] utilizing silver – aluminum oxide
hybrid nanoparticle was carried out. Varying the percent volume
24
concentration from 0.005% to 0.1%. The study reported that a that allow users to create various types of items with defined
grooved heat pipe system with nanofluids and hybrid nanofluids led properties. To create a new type of fluid, properties such density,
to worse system performance. According to a research review [11], dynamic viscosity, specific heat, and thermal conductivity are
there is an incongruity among experimental data of various authors needed at least to define a fluid. Such properties were taken from the
and the issue of heat transfer enhancement or deterioration with calculations made using the equations and correlations which will
nanoparticle utilization is still unsolved and it requires follow up be described in the succeeding sections. Incoming hot water
investigations. Conferring to the review, different data, sometimes temperature, nanofluid inlet temperature, hot water mass flow rate
contradictory ones, are obtained even for the same compositions of and nanofluid mass flow rates will be taken as input parameters in
nanofluids. Numerous available studies and related literature can be the simulation. Such parameters will be based on the results taken
analyzed and cited that shows that there is no quantifiable nor from the experimental aspect of the study.
qualitative unanimity with regards to the heat transfer performance
of nanofluids. Thus, such circumstance requires additional D. Data Reduction
experimental study of the heat transfer of nanofluids. The primary
purpose of this paper – the study on the effect on heat transfer a. Thermo-physical properties
efficiency when adding tungsten carbide (WC) nanoparticles and
cobalt oxide (Co3O4) nanoparticles to distilled water as the base fluid As dispersions of solid particles in a continuous liquid
tested in a shell and tube type crossflow heat exchanger with 0.25- matrix, nanofluids are expected to have a thermal conductivity that
inch inner diameter at laminar and transient regimes. This study also obeys the effective medium theory developed by Maxwell [12] as:
intends to present a computer simulation of the experiment utilizing
SOLIDWORKS Flow Simulation program and assess whether such Knf Knp + 2Kbf + 2Φ(Knp − Kbf )
= (1)
method would prove to be beneficial in finding out the effects of the Kbf Knp + 2Kbf − Φ(Knp − Kbf )
nanofluids investigated.
where Knf is the nanofluid thermal conductivity, K bf is the thermal
conductivity of the base fluid, Knp is the nanoparticle thermal
II. METHODOLOGY conductivity and Φ is the particle volume fraction.
Nanofluids, which are commonly undiluted dispersions of
A. Nanofluid Preparation nanoparticles, are deliberated to be better suitable utilizing the
Nanoparticles investigated in the study were produced by power law-based representations proposed by Krieger and
SAT nano Technology Material Co., Limited. WC nanoparticles Dougherty [13] for the estimation of the viscosity of a nanofluid
were less than 100 nm at 99.95% purity. Co3O4 nanoparticles were paralleled to other representations like the one credited to Einstein
less than 40 nm at 99.5% purity. WC – water and Co3O4 – water [14]. Einstein’s correlation is defined as:
nanofluids were prepared via two-step method. With 0.5% to 2.0%
volume concentration in increments of 0.5%. The mixtures were μnf / μbf = 1 + KμΦ (2)
then placed on a H3770-HS digital magnetic stirrer for 30 minutes
at stirring speed of 1000 rpm. After that, the solution was subjected where μnf is the viscosity of the nanofluid while μbf is the viscosity
to ultrasonic vibration with a Cole – Parmer EW-08895-04 for 1.5 of the base fluid. Kμ = 2.5 as Einstein suggested for dilute systems
hours at 40kHz. The produced nanofluids were stable for at least 4 of spherical particles. Krieger – Dougherty correlation is expressed
hours without any visible sedimentation. as:
Φ - [μ]Φm
B. Test Rig μnf = μbf (1- ) (3)
Φm
25
is the change in temperature. With heat exchangers, the correlation where f is the friction factor relating to friction losses in pipe flow
concerning the overall heat transfer coefficient and the rate of heat and flow in open – channel [24],
transfer can be expressed by the subsequent equation,
f = (0.79ln(Re) – 1.64)-2 (14)
Q = UA (LMTD) (7)
Pe is the nanofluid Peclet number, which is a quantity of
where Q is the heat transfer rate, A is the heat transfer surface area, comparative significance of advection contrasted with diffusion
U is the overall heat transfer coefficient, and LMTD is the [25]. Peclet number can be defined as:
logarithmic mean temperature difference.
For identified values of overall heat transfer coefficient, Pe = Re * Pr (15)
heat transfer coefficient of the nanofluid, hnf, can be determined by
where Re is the Reynolds Number and Pr is the Prandtl Number as
1 defined previously.
hnf = 1 As t A 1 A (8)
( )( )-( )( s )-( )( s )
U Am k Am hs Anf
where As, Am, Anf are shell side surface area, logarithmic mean of
the surface areas, and nanofluid surface area, respectively. t
represents the thickness of the tube, k is the thermal conductivity of
the tube used and hs is the heat transfer coefficient at the shell side,
which can be identified accordingly through [17]
ℎ𝑠 (𝑑𝑜− 𝑑𝑖 ) 𝑑
= 0.020𝑅𝑒 0.8 𝑃𝑟 0.33 ( 𝑜 )0.53 (9)
𝑘 𝑑𝑖
c. Dimensionless Number
26
increasing particle concentration compared to that of the base fluid. introduction of solid nanoparticles which have densities that are
Highest increase recorded at 6.02% for WC-water nanofluid and greater than the base fluid, this would in turn lead to specific heat
5.55% increase for Co3O4-water nanofluid both at 2.0% volume capacity that is lower than water.
concentration. The considered correlation implores the Effective
Medium Approximations (EMA) or Effective Medium Theory
(EMT). Such approximations or estimates are depictions of a
medium centered on its properties as well as the properties of the
qualified fractions of its components. Considering component level
of a certain material, values differ and are diverse. Particular
computation of the several component values is almost irresolvable
but theory as such can come up with satisfactory estimates which
can define valuable parameters and characteristics of the combined
material in its entirety [28, 29]. Also, in a benchmark study [30], it
affirmed that the effective medium approximations or effective
medium theory for well-dispersed particles can accurately
reproduced experimental data in the limited set of nanofluids in their
exercise. With regards to viscosity, as expected viscosities of the
nanofluids are higher than that of the base fluid. Since nanoparticles
are dispersed in the base fluid, it developed its resistance to flow. Fig. 5a. Nanofluid Density
Thus, increasing its viscosity. Fig. 4 shows the graphed results of the
viscosities of the nanofluids compared to approximated results using
different correlations [31 – 33].
27
with Reynolds number in the transient condition (i.e. 2300 < Re < Figures 8a and 8b illustrates the graphed results of the percent
4000). Discrepancy between the different Nusselt number enhancement or detraction of nanofluid with distilled water as the
calculations revealed that the average percent difference between base line.
experimental – theoretical, experimental – simulated, and theoretical
– simulated are 11.793%, 2.775%, and 12.901%, respectively for
WC-water nanofluid. While Co3O4-water nanofluid showed an
average percent difference of 11.364%, 4.937%, and 10.702% for
experimental – theoretical results, experimental – simulated results,
and theoretical – simulated results, respectively. Such observations
of discrepancy between theoretical calculation and experimental
data can also be perceived in a certain study [37]. Wherein
theoretical calculations undershoot experimental results by 50%. A
correlation is then proposed for this study to approximate the results
of the experimental Nu where coefficients of the correlation are
based on percent volume concentration, Reynolds number, Prandtl
Number, and friction factor. The general assessment for the
coefficient of determination on how sound the approximated results
are in the proposed correlation is 97%. Proposed correlations are
described in the concluding section.
Fig. 6d. Co3O4 – Water Nanofluid Nu at Transient Regime
28
IV. CONCLUSION
The current study dealt with inquiries on the utilization of
WC, and Co3O4 nanoparticles on water as the base fluid on its effects
on heat transfer capability. Effects of the considered nanofluids with
percent volume concentrations of 0.5%,1.0%,1.5%, and 2.0% were
analyzed with the main attention on heat transfer efficiency at
laminar to transient regimes in the particular set – up. Considering
the thermo – physical properties, adding nanoparticles to the base
fluid had substantial impacts. Decrease in heat capacity can be
perceived for all the nanofluids. While increase in thermal
conductivity, viscosity, and density was noticed with all nanofluids
investigated. Comparison between theoretical computations,
experimental data, and computer simulated results had significant
differences. Theoretical computations compared with computer
simulated data had the highest discrepancy. While experimental data
Fig. 7c. Co3O4 – Water Nanofluid Simulation and computer simulated results had lesser discrepancy. Regarding
the effects on heat transfer efficiency, WC-water nanofluid
With the largest reduction recorded at 4.89% for 2.0% presented deteriorations in all particle loading even though it
volume concentration. Cobalt Oxide nanofluid had better results at exhibited boost in thermal conductivity for all percent volume
laminar conditions. Recording an enhancement of 1.43% at 0.5% concentration. While Co3O4-water nanofluid proved to be
particle loading, 2.19% enhancement at 1.0% and a high of 4.3% beneficial. With the highest augmentation in heat transfer efficiency
enhancement at 1.5% volume concentration. Then at 2.0% volume of 13.27% at 1.5% volume concentration. Correlations are proposed
concentration, the heat transfer efficiency recorded a reduction of in approximating the Nu of the nanofluids investigated with less than
2.23%. For nanofluids at transient regimes, WC-water nanofluids 10% deviation.
still has not exhibit any significant improvement in heat transfer
efficiency. Nonetheless, the degree of reductions is considerably Nuwc = -17.394 + 30.201Φ + 0.0039Re + 1.26Pr + 284.66f
lower at such regime compared to laminar conditions. For Cobalt
Oxide nanofluid however, it revealed the best outcomes among all valid for 1900 < Re < 3020, 4.9 < Pr < 6.4, 0.5% - 2.0% Φ
nanofluids in both laminar and transient states. Exhibiting a 5.76%
enhancement at 0.5% particle loading, 7.85% enhancement at 1.0% NuCo3O4 = 73.19 – 1260.38Φ + 0.0248Re – 39.53Pr + 2733.203f
particle loading, highest augmentation of 13.27% at 1.5% volume
concentration and 5.92% improvement at 2.0% volume valid for 1900 < R e< 2600, 5.7 < Pr < 6.4, 0.5% - 2.0% Φ
concentration.
ACKNOWLEDGEMENT
REFERENCES
29
[7] Williams W, Buongiorno J, Hu L. Experimental investigation Engineering: R: Reports. 63: 1–30. doi:
of turbulent convective heat transfer and pressure loss of 10.1016/j.mser.2008.07.001
alumina/water and zirconia/ water nanoparticle colloids [30] Buongiorno, Jacopo et al. “A benchmark study on the thermal
(nanofluids) in horizontal tubes. Journal of Heat Transfer 2008; conductivity of nanofluids.” Journal of Applied Physics 106
130:042412. (2009): 094312. Copyright © 2009, American Institute of
[8] Nandy Putra, Wilfried Roetzel, Sarit K. Das, Natural Physics
convection of nano-fluids, Heat and Mass Transfer 39 (2003) [31] Brinkman HC. The viscosity of concentrated suspensions and
775–784. solution. Journal of Chemical Physics 1952; 20:571–81.
[9] Dongsheng Wen; Yulong Ding, Natural convective heat [32] Krieger IM, Dougherty TJ. A mechanism for non-Newtonian
transfer of suspensions of titanium dioxide nanoparticles flow in suspensions of rigid spheres. Transactions of the
(nanofluids), IEEE Transactions on nanotechnology, vol. 5, no. Society of Rheology 1959;3: 137–52
3, May 2006. [33] Batchelor GK. The effect of Brownian motion on the bulk
[10] Han WS, Rhi SH. Thermal characteristics of grooved heat pipe stress in a suspension of spherical particles. Journal of Fluid
with hybrid nanofluids. Therm Sci 2011; 15:195–206. Mechanics 1977; 83:97–117
[11] Terekhov V I, Kalinina S V and Lemanov V V. The mechanism [34] Askeland, Donald R.; Fulay, Pradeep P.; Wright, Wendelin J.
of heat transfer in nanofluids: state of the art (review). Part 2. (2010-06-21). The Science and Engineering of Materials (6th
Convective heat transfer. Thermophysics and Aeromechanics, ed.). Cengage Learning. ISBN 9780495296027
2010, Vol. 17, No. 2. [35] "Stiffness of long fibre composites". University of Cambridge.
[12] J.C. Maxwell, A Treatise on Electricity and Magnetism, second Retrieved 1 January 2013.
ed., Clarendon Press, Oxford, UK, 1881. [36] Halliday, David; Resnick, Robert (2013). Fundamentals of
[13] Krieger IM, Dougherty TJ. A mechanism for non-Newtonian Physics. Wiley. p. 524
flow in suspensions of rigid spheres. Transactions of the [37] Lazarus Godson Asirvatham, Balakrishnan Raja, Dhasan
Society of Rheology 1959; 3:137–52 Mohan Lal, Somchai Wongwises, Convective heat transfer of
[14] Einstein A. Eine neue Bestimmung der Moleküldimensionen. nanofluids with correlations, Particuology 9 (2011) 626–631
Annalen der Physik 1906; 19:289–306.
[15] Pak, B. C. & Cho, Y. I. (1998). Hydrodynamic and heat transfer
study of dispersed fluids with sub-micron metallic oxide
particles. Experimental Heat Transfer, 11, 151–170
[16] Xuan,Y. and Roetzel,W.(2000).Conceptions for heat transfer
correlation of nanofluids. International Journal of Heat and
Mass Transfer, 43, 3701–3707.
[17] Yoo S.J., France D.M., Tarshish T.M.: Post-critical heat flux
swirl flow heat transfer with two refrigerants and water. J.
Thermophys. Heat Tr. 11(1997), 189–195.
[18] Falkovich, Gregory (2018). Fluid Mechanics. Cambridge
University Press. ISBN 978-1-107-12956-6.
[19] Zohuri B., McDaniel P. Thermodynamics in Nuclear Power
Plant Systems. Springer; 2015, ISBN: 978-3-319-13419-2
[20] Coulson, J. M.; Richardson, J. F. (1999). Chemical Engineering
Volume 1 (6th ed.). Elsevier. ISBN 978-0-7506-4444-0.
[21] Wang, Xiang-Qi, & Mujumdar, Arun S.. (2008). A review on
nanofluids - part I: theoretical and numerical investigations.
Brazilian Journal of Chemical Engineering, 25(4), 613-630.
[22] Fang, XD and Bullard, C.W. & Hrnjak, Pega. (2001). Heat
Transfer and Pressure Drop of Gas Coolers. ASHRAE
Transactions. 107. 255-266
[23] Gnielinski, V. (1976). New equations for heat and mass transfer
in turbulent pipe and channel flow. International Chemical
Engineering, 16, 359–368
[24] Munson, B.R. (2006). Fundamentals of Fluid Mechanics (5
ed.). Hoboken, NJ: Wiley & Sons.
[25] Coulson and Richardson's Chemical Engineering (Seventh
Edition) Volume 1b: Heat and Mass Transfer: Fundamentals
and Applications 2018, Pages 529-546 - Transport Processes in
Microfluidic Applications
[26] Hamilton RL, Crosser OK. Thermal conductivity of
heterogeneous two-component systems. Ind Eng Chem
Fundam 1962;1(3):187–91
[27] Maiga SEB, Palm SJ, Nguyen CT, Roy G, Galanis N. Heat
transfer enhancement by using nanofluids in forced convection
flows. Int J Heat Fluid Flow. 2005; 26:530–46
[28] Wenshan, Cai; Shalaev, Vladimir, Optical Metamaterials:
Fundamentals and Applications. November 2009, Springer. pp.
Chapter 2.4. ISBN 978-1-4419-1150-6.
[29] Wang, M; Pan, N. "Predictions of effective physical properties
of complex multiphase materials". 2008, Materials Science and
30
Development of Hybrid Convection Solar Dryer for
Seaweeds (Kappaphycus alvarezii S.)
Reniel Z. Rocaberte1, a *, Marvin M. Cinense2,b, Ruel G. Peneyra3,c
and Jeffrey A. Lavarias4,d
1
Puerto Princesa City, Palawan
234
Central Luzon State University, Science City of Muñoz, Nueva Ecija
a
rocaberte.reniel01@gmail.com, bmarvin_cinense@clsu.edu.ph, crgpeneyra@clsu.edu.ph,
c
jeffreyalavarias@clsu.edu.ph
Abstract—The seaweeds solar drying experiment was needed because inadequate processing and handling of
conducted under metrological condition of Philippines using products can lead to quality loss.
hybrid convection solar dryer with solar thermal air collector and
solar-powered auxiliary heater. This study addressed the problem Drying is an essential agricultural technique used to
encountered by local seaweed farmers in dealing with seaweed preserve food products by reducing the moisture of the
production likewise it based on the need for more efficient and product. Locally, sun drying is one of the conventional
practically cost-effective drying system during cloudy and rainy drying methods that is used by seaweeds farmers, the
season. Seaweeds drying sample was subjected at different drying seaweeds are placed in an open space or a wooden platform
temperature of 40C, 50C and 60C. Under these drying and or hanged outside. However, this method is extremely
conditions the moisture content reduced from 93.70% to 35.31% weather dependent and the material are exposed to foreign
moisture in 440 minutes. It showed that temperature played matters such as soil, sand which susceptible to contamination
significantly to the mass transfer of seaweeds. The drying kinetic infestation, microbial attacks, thus affecting product quality
were fitted to exponential models using the coefficient of (Seveda, 2013). Since the exportation of seaweeds started,
determination (R2) and mean bias error (MBE). Results showed the drying using the conventional method have been used but
that the henderson and pabis model best described the due to intensive utilization in the global market the
characteristics of seaweeds which it was dominated by diffusion. requirement of quality is one of the primary concerns and it
In addition, effect of drying temperature on the color of dried is now requiring improvement (Phang et al, 2014).
seaweeds was evaluated by measuring the RGB value. The image
that was taken during the study was analyzed in capture image Being a weather dependent methods and with the given
software. It reveals that the resulting RGB mean values of dried limitations in open solar drying, there is a clear need to
seaweed at different drying condition were significantly affected develop an innovative dryer addresses the defined issues.
by drying temperature. Overall, this report indicates that drying at Furthermore, such intervention must take advantage of the
60C is most effective with good color appearance attributes with abundant solar energy to ensure low production cost coupled
uniform color. Further, the change of RGB value and moisture at with its capacity to dry during short and reduced sunshine
different drying condition can be used to predict the seaweeds and even at night time to ensure the production of good
drying process. quality dried seaweeds. Though there were several drying
systems have been designed, developed and tested to address
Keywords—color analysis, solar drying, kinetic modeling, ,
the issue such that the seaweeds product will be more
seaweed
competitive and economically viable to venture on. But
expected this innovation requires a lot of energy hence an
I. INTRODUCTION added cost to farmers, reducing their net potential income.
Seaweed is vital for human life because of its nutritional The lack of proper drying facilities and techniques in the
benefits and multiple uses. The Kappaphycus alvarezii S. is intensive quality requirement of dried seaweeds for the
one of the major variables that used to make carrageenan. global demand drives the effort to an innovative solution
This carrageenan known as hydrocolloid which is a natural through enclosed drying system with solar thermal air heat
extract from seaweeds as gelling agent, thickener, and collector and solar-powered backup heater
stabilizer in a wide variety of food applications, non-food
applications and cosmetics, pharmaceutical products. In the
export market, seaweeds are traded as dry to facilitate
efficient transport and product logistic (Phang et al, 2014). II. METHODOLOGY
Consequently, seaweeds must be dried before they can be
utilized for agricultural, industrial, and other purposes. A. Preperation of Sample
Seaweeds are reported as an extremely perishable The seaweeds sample was obtained from the nearest
commodity which in most cases losses its nutritive value seaweeds production area of BFAR- Masinloc, Zambales.
within hours at a tropical temperature and recommended that The harvested sample were wash thoroughly with seawater
quick processing must be undertaken to prevent the losses of to removed epiphytes and other foreign matters. The
the product. Effective preservation techniques, therefore, are seaweeds sample were cut into smaller pieces (4 inches)
before soaking into fresh water. The sample was soaked for
31
about two hours. Then, then initial moisture content of the lowest mean bias error (MBE), root mean square error
seaweeds was determined by oven drying method. The (RMSE), and reduced chi-square (X2) (Ibrahim et al, 2009;
average initial moisture content obtained is 93.70%. The Hao et al, 2020).
dryer was pre-heated for 20 minutes in order to establish
steady state condition to the desire temperature of 40°C, 1 2
MBE= ∑N
i=1(MRpre.i -MRexp.i ) (1)
50°C and 60°C using the temperature controller while the N
1
samples of the same mass (1700kg) were evenly arrange in 1 2 2
the three (3) drying trays. The reduction of weight was RMSE= [ ∑N
i=1(MRpre.i -MRexp.i ) ] (2)
N
determined from each tray at the top, middle and bottom
every two hours. The air temperature and relative humidity D. Color Analysis
inside the dryer and the surroundings were recorded every The dried seaweeds at different temperatures were
ten minutes using a data logger. evaluated using color analyzer. The color analyzer was used
to measure color value of the certain part of an image by
B. Hybrid Convection Solar Dryer zooming into the region of interest and to obtain the RGB
The drying system was fabricated using locally and value. The color change measure in RGB color can be in
commercially available materials with an approved design color channels with intensity value for red, green and blue
plan. The unit and component are in accordance with color from 0-255. The results are presented as the mean
requirement of the seaweed. The dryer has an outer value of RGB (Mendoza et al, 2006; Lukinac et al, 2009).
dimension of 150cm x 150cm x 120cm (length x width x
height) and it was installed in Land and Water Resource E. Statistics Analysis
Management Center (LWRMC) Farm. Figure 1 shows the The data collected in this study were analyzed and
schematic diagram of hybrid convection solar drying interpret using STAR program for agriculture research
system. The fans and auxiliary heater were powered by solar version 2.0 to evaluate the effects of temperature to the
PV system. The auxiliary heater served as back-up resulting performance and quality parameter of dried
mechanism to augment the needed energy for the air to seaweeds. The research design used single factorial
reach its desired temperature for drying. The air temperature experiment arrange in CRD with three replications and
was controlled by a microcontroller-based temperature different drying temperature 40°C, 50°C and 60°C as
controller operating on a feedback mechanism. This was treatments. The test of comparison among means was made
installed to actively regulate the temperature inside by using the Tukey’s Honest Significant Difference (HSD) test.
turning ON/OFF the heater and fan simultaneously during
the drying operation. The dried maximum capacity is 30kg III. RESULTS AND DISCUSSION
per batch. The hybrid convection solar dryer was designed and
fabricated to address the need of the industry for hygienic
and cost-effective solution to seaweeds processing. the
hybrid dryer consisted of a solar thermal air collector as main
heating system with three (3) auxiliary heaters with fans.
That ensure suction and circulation of air from the solar
collector to the drying chamber. This served as the roof of
the dryer and it was installed at 16-degree inclination at the
top of the drying chamber. Both fans and heaters were
powered by solar PV system. The drying chamber was
thermally insulated with polycarbonate sheet while the floor
area was painted with black acting as great heat absorber. A
non-contact temperature sensor made of type-K
thermocouple feeds the temperature level directly to the
controller that triggers the relays to switch the heater
ON/OFF based on the desired set-point.
A. Drying Performance
The temperature inside the dryer shows significant result
with the ambient temperature as shown in the Figure 3. The
Figure 1. Schematic diagram of HCSD of seaweed
temperature inside the dryer shows significant result with the
ambient temperature, where the inside temperature is higher
C. Drying Analysis than the outside temperature. This might be associated with
absorptive effect of the solar thermal air collector and the
Only good quality seaweeds were used in the experiment
auxiliary heater that was properly maintained the heating
5.1kg of fresh seaweeds were dried until the required final requirement of the dryer. The average relative humidity
moisture content was attained. Moisture content was inside the dryer at different drying conditions of 40°C, 50°C
expressed as percent wet basis and then converted to gram and 60°C were 40.17%, 34.07%, and 34.75% respectively.
of water per gram of dry matter. The experimental drying Lower relative humidity was observed at higher temperature
data of seaweed were fitted to the exponential model using due to the effect of heating to the vapor pressure exerted by
non-linear regression analysis. The section of best fit was the vapor in the air (Rathore and Panwar, 2010).
made using the highest coefficient of determination (R2) and
32
Figure 4. Krischer curve-drying rate versus moisture content
The quality of fit was expressed by the highest
coefficient of determination and lowest and the lowest Mean
Bias Error (MBE), Root Mean Square Error (RMSE), and
reduced chi-square (X2) (Fudholi et al 2014). Along with
Figure 2. Hybrid convection solar dryer for seaweed
criteria and the goodness of fit was statistically showed in
Henderson and Pabis model. For all treatments, the 60°C
drying temperature gave lower MBE, Reduce Chi-Square
which corresponds to higher results of R-square value. The
70
results showed a very high percentage of the possibility that
60
Temperature,°C
the model could predict the drying rate behavior at any given
50
condition of drying temperature. This means that Henderson
40
and Pabis model best describes the characteristics of
30 seaweeds in a convection solar dryer where it was dominated
20 by diffusion.
0 50 100 150 200 250 300 350 400 450 500 550 600
Time, min
40°C 50°C 60°C
40°C Amb 50°C Amb 60°C Amb TABLE I. VALUES OF MODEL CONSTANT AND STATISTICAL
PARAMETERS
Figure 3. Variation of drying temperature and ambient temperature
Temperature R2 (%) MBE RMSE X2
Model
Lewis 92.32147 0.40700 1.01736 1.03503
The seaweeds dried at 40°C reached 35% moisture 40°C
content in 590 minutes while seaweeds dried at 50°C and Henderson
86.35699 0.00000 0.25182 0.05888
and Pabis
60°C reached their desired moisture content of 35.78% in
453 minutes and 35.31% within 440 minutes, respectively. Lewis 91.26493 0.48814 1.19031 1.41685
Drying temperature had a great influence to the removal of 50°C
Henderson
surface moisture of seaweed as it was attributed to the 96.22279 0.00000 0.11693 0.01269
and Pabis
highest reduction rate (Yahya 2007). It was observed that
Lewis 91.47816 0.48563 1.18032 1.39316
the increase in drying temperature consequently decreased 60°C
the drying time (Fudholi et_al, 2011). Henderson
99.01455 0.00000 0.06031 0.00337
and Pabis
Furthermore, the external moisture and internal moisture
evaporates at different rate. The external surface of seaweed
had slight changes during the initial adjustment period
wherein the surface moisture was exposed to the drying air
evaporating at constant time, and on the later period, it has
undergone moderate changes in moisture reduction.
Figure 4 shows the relationship of moisture content and
drying rate as represented Krischer curve of dried seaweeds.
It was found that the drying rate initially increased and
reached its maximum value in about two hours from the
beginning of drying operation. This suggests that the rate of
movement and removal of moisture by diffusion and
evaporation in a convection drying process were energy- Figure 4. Plot of experimental and predicted value of 60°C
dependent and requires constant temperature regulation. - Henderson and Pabis model
33
seaweeds to the energy input. The electrical energy input a root mean square error of 0.00108, and reduced chi-square
was measured using the power meter (kWh). The hybrid of 0.000001 at a temperature of 60°C. Thus, Henderson and
convection solar dryer was evaluated considering the Pabis best described the drying characteristics of seaweed
parameter where the heat energy of the solar thermal air drying in a hybrid convection solar dryer. Results of color
collector was supplied by solar radiation and the determination showed that seaweed samples dried at 40°C
supplemental heating system was driven by the power of had the darkest appearance which was attributed to an
PV/T solar panel. %. During the initial drying process, the increase of oxidation reaction.
heater and the fan worked simultaneously to augment the
heat requirement for drying conditions thus, consumed more ACKNOWLEDGMENT
energy when set on high drying temperature. At a The author would like to thank the Department of
temperature of 60°C, both the auxiliary heater and fan had Science and Technology through Engineering Research and
longer ON time with a total of seven hours as compared to Development for Technology (DOST-ERDT) for the
drying temperature of 40°C that had shorter ON time during financial support through the graduate scholarship program,
the drying process with a total of four hours. The functional Central Luzon State University grant scheme. The author is
combination of auxiliary heater and fan was proven to be grateful to Dr. Marvin M. Cinense for the guidance
effective as it gave the best results for the dried seaweeds. throughout the experiment. To Prof Ruel G. Peneyra, his
At 60°C, the seaweeds dried better with lesser drying time second mentor, for overflowing intelligence, knowledge and
of about eight hours but higher rate of energy consumption effort in providing help to completely attain the good
which further resulted to a reduced efficiency with respect research result. Prof Jeffrey A. Lavarias, for his advice and
to the time of operation of the heater and the fan, likewise, fruitful discussions. To Dr. Majid Khan Majahar Ali for the
drying temperature corresponds to the energy utilized. guidance and support. The Bureau of Fisheries and Aquatic
C. Color Quality of dried Seaweed Resources, Zambales Station Mr. Marlon Alejandro for their
support in sending the commodity to the Central Luzon
Color is one of the quality attributes in dried products.
State University in this hard time of pandemic.
This simple color analysis is to compare the difference in
color of the dry seaweeds as affected by the drying REFERENCES
treatments. The results suggest that dried seaweeds at higher
[1] A. Fudholi, M.Y. Othaman, M.H. Ruslan, M. Yahya, A.
temperature had better appearance in terms of color. It is Zaharim and K. Sopian “The effects of drying air
worth mentioning that the temperature at 60°C indicated a temperature and humidity on drying kinetic of
remarkable uniform lightness during drying. The increase of seaweed,” International conference on energy and
value of blue signifies blacken reaction. Temperature at development-environment-biomedicine. 2011, pp. 129-
133.
40°C obtained the higher blue value that signifies the
darkening of color on dried seaweeds. This can be supported [2] A. Mina. 2018. Convection drying of mushroom
(Pleurotus florida) using solar thermal heatpipe water
by the observation on dried commodity that lower heater. Master Degree Thesis. Central Luzon State
temperature setting results to darker colored surface. Similar University.
results were observed in the study of Mina, (2018) on which [3] A. U. Sundaria, P. Neelamegamb and C.V.
the browning of the surface of the dried product is due to Subramanianc, “Drying kinetics of muscat grapes in a
effect of lower drying temperature and thus significantly solar dryer with evacuated tube collector. International
journal of Engineering,” 2014, 27(5):811-818.
affects the final product. [4] BFAR, “Fisheries Commodities Road Map,” Seaweeds:
Fisheries Policy and Economic Diversion; Bureau of
IV. CONCLUSION Fisheries and Aquatic Resources. 2010.
In this study, to help seaweed farmers produce [5] D. G. Mercer. 2008. Solar drying in developing
countries: possibilities and pitfalls. International Union
white dried seaweeds anytime, a hybrid convection solar of Food Science and Technology. 4:1-11.
dryer was designed and fabricated using locally available [6] D. J. McHugh, “A guide to the seaweed industry,” FAO
materials. It is equipped with heater and solar thermal air Fisheries Technical Paper 441. Rome: Food and
collector, and it utilizes heat from the sun. The solar thermal Agriculture Organization, 2003.
air collector combined with solar-powered auxiliary heater [7] D.Y. Dasin, A.I. Zakari and I. Yahuza, “Development
provide enough hot air to dry the seaweeds. Likewise, the of solar energy dryer for fish and meat tested in
Modibbo Adama University of Technology Yola,”
use of temperature controller maximized the utilization of International Journal of Trend in Scientific Research
heat in the system and it gave a satisfactory result in and Development. 2019, 3(2):pp. 774-777.
maintaining and regulating the temperature of air inside the [8] F. Mendoza, P. Dejmek and J. Aguiler, Calibrated color
drying chamber between temperature levels of 40°C, 50°C measurements of agricultural foods using image
analysis. Postharvest Biology and Technology. 2006,
and 60°C.The moisture content of seaweeds in the hybrid 41(3):285-295.
convection solar dryer was reduced from an initial moisture [9] H. K. Phang, C. M. Chu, S. Kumaresan, M. Mizanur
value of 93.70% (w.b) to a final value of 35% (w.b) within Rahman and S. M. Yasir. “Preliminary study of
ten hours of drying. The drying temperatures of 60°C seaweed drying under a shade and in a natural draft
significantly reduced the moisture and increased drying rate solar dryer,” International Journal of Science and
Engineering, 2014, 8(1):10-14.
thus, shorter drying time compared to 40°C. However, lower
[10] J. Lukinac, S. Jokic, M. Planinic, D. Magdic, M. Bilic,
temperature level resulted to lower energy consumption. S. Tomas, D. Velic and A. Bucic-Kojic, “ An
Consequently, The Henderson and Pabis model showed the application of image analysis and colorimetric methods
best fit with a high value of 99.01% for the coefficient of on color change of dehydrated asparagus (Asparagus
correlation and lower values of mean bias error of 0.00005,
34
maritimus L.),” Agriculturae Conspectus Scientificus.
2009, 74(3): pp. 233-237
[11] M. K. M. Ali, A. Fudholi, J. Salaima, M.S. Muthuyala,
S. M. Yasir, A. Q. Hortado, “Post-harvest handling of
Eucheumatoid Seaweeds: Tropical Seaweed farming
Trends, Problems and Opportunity: Focus on
Kappaphycus and Eucheuma of Commerce,” 2017, pp.
131-145.
[12] N. Rathore and N. Panwar, “Design and development of
energy efficient solar tunnel dryer for industrial drying
Clean Technology,” 2010, Environment. 13 pp 125-32
[13] S. Suherman, M. Djaeni, A. Kumoro, A. Prabowo, S.
Rahayu and S. Khasanah. “Comparison drying behavior
of seaweed in solar, sun and oven tray dryers,” 2018,
MATEC conference 156.
[14] S. Abasi,S.M. Mousavi, M. Mohebi and S. Kiani,
“Effect of time and temperature on the moisture
content, shrinkage, and rehydration of dried onion,”
Iranian Journal of chemical Engineering. 6(3): 2009, pp.
58-70.
[15] P. Sain, V. Songara, H. Balan and R. Karir. “Natural
convection type solar dryer with latent heat storage,”
2016,
https://www.researchgate.net/publication/286662466.
[16] W. Hao, S. Liu, B. Mi and Y. Lai, “Mathematical
modeling and performance Analysis of new hybrid
solar dryer of lemon slice for controlling drying
temperature,” Energies, 2020, pp. 13:350
[17] N. Rathore and N. Panwar, “Design and development of
energy efficient solar tunnel
[18] dryer for industrial drying Clean Technology,” 2010,
Environment. 13 pp 125-32
35
GIS-based Site Suitability Analysis for Biomass
Energy Plant in Bohol’s Rice and Coconut
Producer Municipalities
Wenyville Nabor Galang1,2, a *, Ian Dominic Tabañag3,b and Michael Loretero4,c
1
School of Engineering, University of San Carlos, Cebu City, Cebu
2
College of Engineering and Computer Studies, Holy Name University, Tagbilaran City, Bohol
3
Department of Chemical Engineering, University of San Carlos, Cebu City, Cebu
4
Department of Mechanical and Manufacturing Engineering, University of San Carlos, Cebu City, Cebu
a
wgalang@hnu.edu.ph, biftabanag@usc.edu.ph, cmeloretero@usc.edu.ph
Abstract—Site suitability is an essential process in determining households because of its abundance, accessibility and
potential sites for power plant development with respect to some affordability, particularly in the rural areas [4]. Biomass
limiting factors. This study utilized a Geographic Information gradually stepped up its contribution to the power sector as
System (GIS)-based land suitability analysis model considering its level of fuel input to electricity generation increased by
accessibility for locating optimal sites to the biomass power plant
development against environmental and economic pull factor. In
as much twice its 2015-2016 level.
order to determine suitable site, the important factors are The province of Bohol in Central Visayas is a main
incorporated, namely accessibility and topography, land cover, and producer of agricultural products in the region. It is called
economic by using multi-criteria decision analysis (MCDA)
the “Rice Granary of Central Visayas” after it was cited for
method in decision making process. The integration of MCDA and
a GIS model, allowed the identification and analysis of optimal
surpassing other provinces in the region in terms of rice
suitable locations that balanced economic and environmental output [5]. The rice production in the province hiked by
criteria within the biomass supply. The study produced a suitability 54.2 percent from 161,003 metric tons in 2016 to 248,314
map showing the potential sites for biomass power plant metric tons in 2017, representing the bulk of the rice
development in the area of consideration. It was found that among produced in the region. The province took the bigger share
the top rice and coconut areas in the province of Bohol, the most of the rice production in Central Visayas, improving by 45
suitable site is the municipality of Ubay. percent from 231,982 metric tons in 2016 to 335,420 metric
tons in 2017. In addition, the province is abundance in
Keywords—Biomass, Waste to Energy, Energy Potential, Site
coconut plantation. Bohol Island has around 10 million
Suitability, GIS-MCDA
coconut trees producing an annual harvest of more than 343
I. INTRODUCTION million coconuts [6].
Bohol is primarily supplied by power plants located in
Globally, the use of renewable energy as an alternative Leyte and Cebu through the Leyte-Bohol Transmission
to fossil fuel is highly considered to counter the increasing Interconnection. Power plants outside Bohol supply about
fuel cost and environmental concerns. Biomass is 68.86 MW or 89% of the peak demand of the province
considered as one of the main option for renewable energy through the Leyte-Bohol Interconnection. The remaining 8.4
sources given its abundance to nature [1]. MW (11%) is collectively supplied by the three mini-hydro
plants inside Bohol [7]. The vulnerability of its electric
The term biomass refers to any organic material derived power system and the need to make it resilient to the effects
from plants. Biomass is produced by green plants that of natural disasters lead to recognize the need to put in place
convert sunlight into plant material through photosynthesis. measures that can ensure sufficient power is supplied to the
Biomass resource can be considered as an organic material, whole island-province in the event that it is disconnected
in which the solar energy stored in chemical bonds. When from the rest of the Visayas Grid. Given the abundance of
the bonds between adjacent carbon, hydrogen and oxygen agricultural sources, the province has the means of potential
molecules are broken by digestion, combustion, or energy from agricultural products waste such as rice and
decomposition, then it releases their stored chemical energy coconut, respectively.
[2]. Biomass basically can be divided into three major
groups, wood biomass, non-wood biomass and secondary However, the meaning of potential is not just defined by
fuels [3]. energy transfer, but defined by economic and environmental
factors as well. The location of a bioenergy plant is affected
In the Philippines, biomass as a renewable energy by a wide variety of factors and criteria. This study seek to
resources have great energy potential considering its identify the most suitable locations for a renewable energy
abundance in the sorrounding of an agricultural country. It plant in the identified top rice and coconut municipalities in
represents 25.5% of the indigenous energy supply as of the province. To find the best suitable land for bioenergy
2016. Bulk of the demand for biomass was consumed by the
household sector. It is the most preferred fuel among
36
facility based on multiple criteria decion analysis (MCDA) municipalities of Bien Unido, Buenavista, Clarin, Dagohoy,
encorporated in a Geographical Informational System (GIS). Danao, Getafe, Inabanga, Pres. Carlos P. Garcia, Sagbayan,
San Isidro, San Miguel, Talibon, Trinidad and Ubay; and
The GIS support the solution of complex spatial
the rest of the municipalities particularly Alicia, Anda,
problems, providing the decision-maker with a flexible
Batuan, Bilar, Candijay, Carmen, Dimiao, Duero, Garcia
environment in the process of the decision research and in
Hernandez, Guidulman, Jagna, Lila, Loboc, Mabini, Pilar,
the solution of the problem. To create visualized suitability
Sevilla, Sierra Bullones and Valencia belong to the 3rd
map for users and decision makers, the integration of
district.
MCDA and GIS has been widely promoted for solving
spatial problems in urban assessment and planning [8].
The availability of biomass residues varies spatially in the
Another study was conducted to introduce the Multi-criteria
province of Bohol, mostly due to land cultivation area and
decision analysis into the land use suitability analysis along
land topography. First district of Bohol have 15
with the existing perspective evolving the role of GIS [9].
municipalities including the city of Tagbilaran. The second
Meanwhile, some case studies were performed using the
district is abundant in rice and coconut production. It
method of integrating GIS and multi-criteria analysis to
includes the municipalities of Bien Unido, Buenavista,
create an output with a set of land use suitability maps
Clarin, Dagohoy, Danao, Getafe, Inabanga, Pres. Carlos
incorporating complex criteria integrating several stake-
Garcia, Sagbayan, San Isidro, San Miguel, Talibon,
holders' points of view. In the Philippine setting, a study
Trinidad and Ubay. The rest of the 19 municipalities belong
using Multi Criteria Decision Analysis and GIS processes
to the third district which is abundant in corn production.
were employed in the province of Nueva Ecija. The land
cover was classified from LandSAT 8 image using Support
Vector Machine classification technique. The resulting
classified map was used to generate the biomass resource
map along with expert consultation for a three level
hierarchy of factors used in the site suitability analysis [9].
However, in the region of Visayas, there is limited study
done for site suitability particularly in the province of Bohol
given its biomass potential for renewable energy sources.
This study utilizes the integration of GIS-MCDA to support
decision-makers in providing them with ways to evaluate
several alternatives, based on multiple and conflicting
criteria.
II. METHODOLOGY
The study was conducted within the scope of Bohol Fig. 2. Bohol’s Rice and Coconut Producer Municipalities
province in the country’s Central Visayas region. There are
3 districts composed by 47 municipalities and a city in To determine a suitable land for a new plant, the
Bohol with a total land area of 411,726 ha and arable area of following major approach was developed: (1) selection of
256,400 ha., see Fig. 1 cultivation area (2) ranking and (3) determine the most
suitable land.
The initial step where cultivation areas were selected based
on multiple criteria:
1. road accessibility,
2. slope of the area,
3. large cultivation area;
4. populated area.
37
The input data are vectors in shape file processed using
QuantumGIS, a free and open-source cross-platform
desktop application that supports viewing, editing, and
analysis of geospatial data [10]. The QGIS steps involve
creating a new field in the attribute table of each layer added
for a particular vector and classify them according to
importance. The new field shall be used as an input to
rasterize the respective layers for major road, slope, land
area and population.
Fig. 6. Population
38
QGIS following the waightage given for each criterion. The School of Engineering, University of San Carlos, Cebu City,
satellite view of the specific location is given in Figure 8. Cebu, Philippines;
College of Engineering and Computer Studies, Holy Name
University, Tagbilaran City, Bohol. Philippines
2.Dr. Ian Dominic Tabañag
Department of Chemical Engineering, University of San
Carlos, Cebu City, Cebu, Philippines
3.Dr. Michael Loretero
Department of Mechanical and Manufacturing Engineering,
University of San Carlos, Cebu City, Cebu, Philippines
ACKNOWLEDGMENT
REFERENCES
[1] Woo, H., Acuna, W., Moroni, M., Taskhiri, M.S., and Turner,
P., 2018. Optimizing the Location of Biomass Energy
Facilities by Integrating Multi-Criteria Analysis (MCA) and
Geographical Information Systems (GIS). MDPI Forest
[2] McKendry P. , 2002. Energy production from biomass (part
1): overview of biomass. J. Biores Tech, 83, 37–46
Fig. 8. Biomass Plant Suitability Map in Satellite View [3] Calle F, Rosillo P, Groot S L, Hemstock and J Woods, 2007.
The Biomass Assessment Handbook: Bioenergy for a
sustainable environment. London: Earthscan
[4] Department of Energy, 2017. Energy Demand and Supply
IV. CONCLUSION Outlook 2017- 2040
[5] Philippine News Agency. Available from:
Suitability maps as methods for visualization problem https://www.pna.gov.ph/articles/1024700 [Accessed April 7,
2020].
provided by the combination of MCDA and GIS methods
can therefore be seen as a powerful tool for solving power [6] Biological Homeopathic Medical Corp.. Available
planning problems, such as the location of biomass plants. from:
GIS-MCDA techniques can used to obtain territorial https://bhmcproducts.com/p5-m-coconut-hub-to-rise-in-
information for planning power supplies. In addition it can bohol/?v=a25496ebf095 [Accessed April 7, 2020].
provide the necessary tools to integrate this knowledge into [7] Province of Bohol, 2015. Bohol Island Power Supply Plan
the development of the project to support decision making (2016-2045)
and guarantee sustainable activities. [8] Phua M., Minowa, M., 2005. A GIS-based multi-
GIS-MCDA model as a decision support system, allows criteria decision making approach to forest conservation
efficient resolution of complex problem such is plant site planning at a landscape scale: a case study in the
selection. The development of the method lead to the Kinabalu Area, Sabah, Malaysia. Land Scape Urban
selection of the most suitable site for the biomass plant Plan 71(2), 207–222.
location within the areas with higher value with respect to [9] Saladaga, I., Remolador, M., Sevilla,K., Baltazar, B.,
evaluation criteria (weighted criteria). Inocencio, L.C., Ang, M.R. 2015. Site suitability analysis for
biomass power plant development in nueva ecija, philippines
using landsat based biomass resource map. Environmental
A. Authors and Affiliations Science
1.Wenyville Nabor Galang [10] QGIS Official Website. QGIS. Retrieved 16 October 2013
39
[11] Chen, J.. 2014. GIS-based multi-criteria analysis for [13] Department of Energy, 2018. Available from:
land use suitability assessment in City of Regina. https://www.doe.gov.ph/sites/default/files/pdf/e_ipo/02_eib_r
Environmental Systems Research egion_vi_visayas_power_situation_and_outlook.pdf
[Accessed August 19, 2019]
40
Environment and
Infrastructure
Session Chair : Dr. Sheila K. Ramos
Moderator : Dr. Bryan G. Alamani
Panelists : Dr. Mary Donnabelle L. Balela
Dr. Candy C. Mercado
Dr. Annaliza Rollon
Dr. Maria Antonia Tanchuling
Colorimetric Detection of Copper Ions in Water using
Polymer-decorated Silver Nanoparticles
Edgar Clyde R. Lopez1, a, *, Jeanette Vivien R. Jalalon1, b, Mary Faith C. Delas Alas1, c, Cheska Mae P. Aldana1, d,
Marlon L. Mopon11, e, Jem Valerie D. Perez1, f, *
1
Nanotechnology Research Laboratory, Department of Chemical Engineering, University of the Philippines Diliman
a
edgarclydelopez09@gmail.com, bjrjalalon2@up.edu.ph, cmcdelasalas1@up.edu.ph, dcpaldana@up.edu.ph,
e
mlmopon@up.edu.ph, f jdperez@up.edu.ph
42
concentrations were observed in Figure 2(d). In particular, the
absorbance intensity of the peak at λ = 419.8 nm and λ = 617.7 nm
increases as the Cu (II) concentration increases. As such, the ratio
between these two peaks (617.7 nm / 419.8 nm) can be used as a
signal for detecting Cu (II) concentrations. A calibration curve was
generated based on the absorbance ratio and the Cu (II)
concentration and is shown in Figure 2(e). A strong linear
correlation was established between the absorbance and Cu (II)
concentrations within the range of 0.10 – 4.00 mM Cu (II). Our
generated calibration curve can be used for determining the
concentration of Cu (II) in an unknown water sample at a 95%
confidence level.
We also investigated the selectivity of the PD-AgNPs
against highly-valent cations such as Cr6+, Fe3+, and Al3+, divalent
cations such as Sn2+, Fe2+, Ni2+, Pb2+, Mn2+, Cd2+, Zn2+, Ca2+, Ba2+,
and Ca2+, and monovalent cations such as K+ and Na+. Common
Figure 1. Response contour plots of PD-AgNPs showing: (a) effect metal cations present in various water bodies show only a tiny
of CS and GLA (PEI = 1.75 wt.%) and (b) effect of CS and PEI response relative to Cu (II), as shown in Figure 2(e). As such, these
(GLA = 1.75 wt.%) on 𝛷; (c) effect of CS and GLA (PEI = 1.75 metals will not interfere with sensing Cu (II) ions in water.
wt.%), (d) effect of CS and PEI (GLA = 1.75 wt.%) and (e) effect of
GLA and PEI (CS = 1.00 wt.%) on LOD. IV. CONCLUSION
(minimum LOD). The following formulation of PD-AgNPs is In this study, a reduced quadratic model was shown to fit
deemed optimum: 1.1642 wt.% CS, 0.8298 wt.% GLA, and 1.3687 the experimental values accurately obtained using both the 𝛷 and
wt.% PEI. Our models predicted that the optimized PD-AgNPs has LOD as responses with 95% confidence. Our optimized nanosensor
𝛷 value of 82.5589 ± 0.6907 and LOD at 2.1566 ± 0.2243 mg L-1. composition is comprised of 1.1642 wt.% CS, 0.8298 wt.% GLA,
These values agree with our validation experiments (actual 𝛷 = and 1.3687 wt.% PEI. The optimized PD-AgNPs are monodisperse
82.9540 ± 0.1772 and LOD = 2.2498 ± 0.0597 mg L-1) at a 95% with an average diameter of 15.66 nm and peak absorbance at 419.8
confidence level. This shows that our models are robust and can nm. CS and PEI coordinate well with copper ions, which cause
generally be applied for predicting 𝛷 and LOD values of other distinct changes in the optical spectra of AgNPs. A strong linear
concentration combinations within the design range investigated. correlation was established between the absorbance ratio and the Cu
The optimized PD-AgNPs show a deep-yellow color and are by an (II) concentrations within the range of 0.10 – 4.00 mM Cu (II) with
optical spectrum exhibiting an LSPR peak at λ = 419.8 nm as shown a detection limit of 2.2498 mg L-1. PD-AgNPs were found to be
in Figure 2(a) and an average FWHM of 63.3641 ± 0.1134 nm, selective towards copper compared to other metals commonly found
indicating a narrow particle size distribution. The PD-AgNPs have in water. Here we showed that our optimized PD-AgNPs could be
an average diameter of 15.6617 ± 1.3418 nm, which is typical of used for the rapid and facile colorimetric detection of Cu (II) in
pseudo-spherical silver nanoparticles [3]. water, thus making on-site copper detection a reality.
REFERENCES
[1] B. O. Anyanwu, A. N. Ezejiofor, Z. N. Igweze, and O. E.
Orisakwe, "Heavy metal mixture exposure and effects in developing
nations: An update," Toxics, vol. 6, no. 4, p. 65, 2018.
[2] M. Mahmoudpour, J. Ezzati Nazhad Dolatabadi, M. Torbati, and
Figure 2. (a) Vis-NIR spectra of bare PD-AgNPs; (b) TEM image A. Homayouni-Rad, "Nanomaterials based surface plasmon
of bare PD-AgNPs; (c) TEM image of PD-AgNPs after addition of resonance signal enhancement for detection of environmental
2.0 mM Cu2+; (d) Vis-NIR absorption spectra of the optimized PD- pollutions," Biosens. Bioelectron., vol. 127, pp. 72–84, 2019.
AgNPs at varying Cu2+ concentration; (e) calibration curve for Cu2+ [3] B. Khodashenas and H. R. Ghorbani, "Synthesis of silver
sensing using PD-AgNPs; (f) selectivity of PD-AgNPs against other nanoparticles with different shapes," Arab. J. Chem., vol. 12, no. 8,
metal ions in water. pp. 1823–1838, 2019.
[4] L. Tang and J. Li, "Plasmon-based colorimetric nanosensors for
Upon addition of Cu (II) to PD-AgNPs, the AgNPs ultrasensitive molecular diagnostics," ACS Sensors, vol. 2, no. 7, pp.
changed in color from golden-yellow to green, as seen in Figure 857–875, 2017.
2(a). This may be due to the complexation reaction between PEI and
Cu (II). This caused the aggregation of the individual PD-AgNPs, as
evidenced by the TEM images in Figure 2(b-c). Strong changes in
the optical absorbance spectra of the PD-AgNPs at varying Cu (II)
43
Geotechnical Characterization of Quarry Waste
Limestone as an Alternative Road Base Course
Material
Emerzon S. Torres1, a *and Mary Ann Q. Adajar2,b
1
Graduate Student, De La Salle University, Taft Avenue, Manila
2
Professor, De La Salle University, Taft Avenue, Manila
a
emerzon_torres@dlsu.edu.ph, bmary.ann.adajar@dlsu.edu.ph
Abstract — The rapid infrastructure development in the providing more sustainable road construction materials. The
Philippines over the past years poses problem on construction cost objective of this study is to characterize the quarry waste
due to the limited supply of suitable road foundation materials limestone in a quarry site in Norzagaray, Bulacan, Philippines.
within economic haul distances. An economic and sustainable This study aims contribute to answering the problems of the
solution is to use locally available but suitable materials. This study scarcity of locally-available construction materials and quarry
focused on geotechnical characterization of quarry waste Bulacan waste management and disposal.
limestone as an alternative road base course material. It includes
physical characterization using ASTM Standards, material
composition using petrographic analysis and SEM-EDX image II. METHODOLOGY
identification and bearing strength through California Bearing Ratio
(CBR) test. The results of the study showed that the quarry waste A. Source of quarry waste limestone
Bulacan limestone is a potential road base course material and pilot The quarry waste limestone was obtained from a quarry
testing is highly recommended to perform. site in Norzagaray, Bulacan as shown in the figure below. A
very wide range of particles was found in the site. Washing,
Keywords—limestone, road base course, CBR, petrographic drying, and sieving were performed before the start of each
analysis, SEM-EDX
experiment. Manual crushing was done to meet the gradation
needed for some tests.
I. INTRODUCTION
The "Build, Build, Build" program of the Department of
Public Works and Highways (DPWH) of the Philippines
promises a significant construction development. In the recent
Philippine Infrastructure Report of 2017, strong growth in the
construction and infrastructure industry was predicted to
occur over the next five years [1]. It follows that there is a
greater need for construction materials such as embankment
a. quarry site in Bulacan b. washed limestone
and road foundation aggregates. Searching for alternative
construction materials is considered a sustainable way of Figure 1. Limestone used in this study
development if secondary material is free from hazardous
components and does not have any adverse long-term effects. B. Experimental program
Road construction costs can be very high because of the The laboratory tests were performed on the samples. These
limited opportunity of locating suitable road base material procedures were based from either ASTM or AASHTO
within economic haul distances. It is preferable and standards to establish a reference in evaluating the index
economical to use and to stabilize, if necessary, locally properties and strength of the alluvial soil. Morphology
available but abundant materials than to import conventional includes the SEM-EDX (Scanning Electron Microscope –
aggregates [2]. Energy Dispersive X-ray) analyses, and petrographic analysis.
The 2017 Minerals Yearbook shows that limestone quarry These material analyses supported the results of laboratory
production in 2014 reached 77, 665 thousand metric tons, and tests and explained the soil behavior that is important for long-
it is the third top quarried industrial mineral commodity in the term applications.
country [3]. Its production rate doubled since 2010. The trend CBR tests in accordance with AASHTO T193 were
in quarrying increased the production of limestone waste performed for the limestone aggregates compacted at the
material as well. The abundance of these crushed limestone maximum dry density and under soaked conditions to simulate
wastes creates an opportunity to use it as an embankment the worst possible scenario in the field. Moreover, specimens
material or aggregate replacement. were soaked for 96 hours (4 days). However, a shorter
Sustainability calls on the use of alternative locally- immersion period of not less than 24 hours was applied, as
available materials and waste reduction in our quarry sites. suggested by the AASHTO standards, if the shorter period did
The utilization of waste limestone can be a big help in not affect the test results.
44
III. RESULTS AND DISCUSSION the petrographic analysis that calcite with chemical formula
CaCO3 comprises the majority material.
A. Geotechnical laboratory test results
Using the Grading A of the DPWH Blue Book [4] the
results of the experiments are summarized in Table 1. The
value of 34% shows a very strong resistance to abrasion of the
Norzagaray limestone. The range of values for the Los
Angeles abrasion value of Cebu limestone is 42 – 61% [2].
The ratio of particles passing No. 200 and particles passing
No. 40 or the dust ratio and the Atterberg limits were within (a) (b)
the range of values for road base course application. For the
Figure 2. SEM micrographs of alluvial soil finer than .075mm at (a) x5000
bearing strength, the CBR values of 225 and 340 for 2.54 and (5𝜇), and (b) x7500 (1𝜇) magnification.
5.08mm penetrations surpassed the stipulated 80% soaked
CBR value for road base course materials.
IV. CONCLUSION
The strong resistance of limestone aggregates to bearing
After laboratory tests and analyses on the suitability of
are attributed to the angular shape of the particles, the
Norzagaray quarry waste limestone as a road base course
cementitious property of limestone, and its high resistance to
material were performed, the following conclusions were
abrasion. The results of the morphology and petrography of
made:
the limestone explain this further.
1. the coarse aggregates of limestone has an abrasion value
TABLE I. MATERIAL CONSTANTS
of 34% and the fine aggregates has no plasticity;
DPWH Specs 2. the bearing strength in terms of CBR are 225% and
Soil Property This study
Grading A or B
340% for 2.54mm and 5.08mm penetrations, respectively;
Particle Size Distribution Grading A
Abrasion Value Not greater than 50% 34% 3. the limestone is a biomicrite with 65% of
Dust ratio ≤ 66% 0.5% microcrystalline calcite matrix; the results of SEM-EDX and
Apparent Specific Gravity - petrographic analysis supported the positive findings of the
(coarse) 2.6609 geotechnical laboratory tests;
Specific Gravity (fine) - 2.5375
Liquid Limit, LL % Not greater than 25 Non-plastic 4. the Norzagaray quarry waste limestone is a suitable
Plastic Limit, PL % Not greater than 6 Non-plastic roadbase course material as per the DPWH specifications.
Min. Void Ratio, emin - 0.28806
Max. Void Ratio, emax - 0.51798 Pilot testing is warranted to determine the long-term
Max. Dry Unit Weight,γdmax - performance of the road base course limestone aggregates. It
kN/m3 20.25386 is also recommended to observe the weathering and possible
Optimum Moisture Content, - dissolution that may happen when calcium carbonate reacts
ωopt % 7.07014 with carbon dioxide and water. The effects on the wear of tires
CBR at 2.54mm penetration, % 80 225.29
is suggested for the application of limestone to pavements.
CBR at 5.08mm penetration, % 80 340.39
ACKNOWLEDGMENT
B. SEM-EDX and Petrographic analysis The authors would like to acknowledge the support of
Petrographic analysis was performed in this study to DOST-ERDT for providing financial assistance for this
determine the chemical and physical characteristics of a rock research, the laboratory testing assistance of DPWH-BRS, and
sample. It was employed to identify the rock name as well. the help of Mr. Michael Torres for the material acquisition.
The analysis was also conducted by the geologists of the Also, the study would not be possible without the help of the
DENR-MGB in their geological laboratory services section. DLSU Civil Engineering Faculty.
The limestone is composed of 30% carbonate fragments,
REFERENCES
65% of microcrystalline calcite, and 5% microsparite which
acts as a cement. It was named biomicrite after Folk’s
Limestone Classification [5]. This limestone composition [1] Philippines Infrastructure Report Q3, 2017,
supports the positive results of the laboratory tests on abrasion https://www.giiresearch.com/report/bmi177867-
infrastructure.html.
and CBR.
[2] DPWH, “Use of weathered limestone (Anapog) as blending
The micrographs of the limestone sample in Fig. 2 materials for base course (Item 201). Manila: Department of
revealed clues that it is composed of grains of different shapes Public Works and Highways”, 1991.
and sizes from different materials. Highly visible are the [3] Fong-Sam, Y., “The mineral industry of the Philippines.
Minerals Yearbook 2013: Area Reports: International, Asia
particles that are intact glued together in the matrix. and the Pacific, 2”, 2017.
The EDX revealed that the limestone consists of 52.96% [4] Department of Public Works and Highways, “Standard
oxygen, 38.79% carbon, and 7.21% calcium with traces of specification for Highways, Bridges and Airports”, 2013.
0.62% silicon and 0.42% magnesium. This is consistent with [5] Folk, R.L., “Stages of textural maturity in sedimentary rocks.
Journal of Sedimentary Research”, 21 (3), 127 – 130, 1951.
45
A nonlinear optimization model for subsidy allocation
to disrupted infrastructures and its application in the
Philippines during the COVID-19 pandemic 1
Abstract— There have been several extensions to the levels to increase infrastructure resilience and speed up their
classical Leontief Input-Output model applied in infrastructure risk recovery from disruptions – leading to a reduction in economic loss.
research. Among the said extensions is the Inventory Dynamic
Inoperability Input-Output Model (I-DIIM). The I-DIIM illustrated While the macroeconomic contributions of inventory have
the importance of inventory to the preparedness and resilience of been widely recognized in recent literature, it should be noted that
infrastructures in a production environment, e.g., a macroeconomic inventory is something to be minimized for for-profit institutions
system, to disruptive events. Although the economic implications of rather than promote because of its associated costs. Furthermore, in
such an event to a macroeconomic system are primarily the regular times, the likelihood of calamities (e.g., pandemic) to occur
government's concern, no work has yet attempted to provide a way is very low, which further demotivates businesses to keep inventory
to empirically determine the optimal intervention required from just to prepare for such unlikely events. However, the significance
governments to facilitate the preparedness and resilience of of inventory may be of more value to decision-makers concerned
economic sectors. This work addresses this significant gap in the with the national economy. Since a government aims to maintain a
literature by developing a Subsidy Optimization Model (SOM) sustainable national economy, it is, therefore, in its best interest to
based on the I-DIIM. This work also proposes a simulation-based support economic sectors in crisis. While governments' role in
subsidy optimization procedure with the Generalized Reduced supporting enterprises during crises is an obvious fact, no attempt
Gradient (GRG) with a Multi-Start algorithm for solving the SOM. has yet been made in previous literature to explicitly measure the
The SOM's applicability is tested through an illustrative case study optimal intervention required from governments to minimize
in the Philippines during the COVID-19 pandemic. This work economic loss consequent to calamities. Furthermore, such analysis
revealed that the SOM differs from conventional methods for is also not yet explored while considering the complex
allocating divisible resources at low budgets. The analysis also interdependencies of these sectors. Such analysis is vital in making
highlighted the SOM's ability to exploit and leverage economic these decisions because changes made in one of the macroeconomic
sectors' interdependency to develop an allocation scheme that system's elements, may it be through some policy, may cause a
minimizes total economic loss, which would otherwise be an discreet yet critical impact to the entire system because of their
incredibly difficult task to fulfill without it. The robustness of such interconnectedness.
results is tested through a sensitivity analysis. Some practical
implications of the findings in the Philippines are detailed in this A financial subsidy is among the most common
work. government intervention to support economic sectors in crisis [5].
The financial subsidy helps the sectors sustain their operations while
Keywords— Infrastructure Risk Analysis, Input-Output in a disrupted state. In line with this, it can be assumed that financial
Economics, Optimization, COVID-19, Disaster Preparedness subsidy converts into some inventory upon allocating these sectors.
Inventory is defined herewith as an asset that can be kept and used
I. INTRODUCTION to maintain a sector's ability to provide goods or services to other
infrastructures, sectors, and final consumers while it is in a disrupted
A. The Role of Inventory in Infrastructure Risk Management state. For instance, inventory in the oil and gas sector could include
excess stored gasoline ready for use [2]. In the electric power sector,
The focus of infrastructure risk management and decision- backup generators could provide some total output to be maintained
making has recently shifted from preventing and protecting [1]. Note that these sectors are interdependent – a sectors' output is
infrastructures from calamities to recovery and response [1]. an input to another – which has several economic implications [4],
Inventory availability contributes significantly to infrastructures' as previously discussed. The intrinsic interdependencies among
preparedness and resilience [1-3]. The efficacy of inventory as a these sectors can trigger the propagation of disaster consequences
mitigation strategy to delay the impact of disruptive events has been that often result in a broader range of inoperability and amplified
widely explored in previous literature, e.g., [1], [3-4]. For instance, losses [3]. Thus, assessing these economic sectors' ability to
[3] showed that an optimal sectoral inventory configuration could overcome shocks is an essential factor in allocating financial subsidy
minimize economic loss for a given disruption scenario. Some to configure inventory levels and optimize their preparedness for
works also explored the impact of inventory in real cases. For critical events; to reduce damage, ensure resilience, and mitigate
example, [4] illustrated the importance of inventory in reducing monetary losses [6-7].
Philippine economic sectors' inoperability during the typhoon
Yolanda (internationally known as Haiyan). These works provided
strong empirical results supporting the need to configure inventory
The present work was part of the master's thesis of the author funded by Engineering Research and Development for Technology (ERDT) scholarship of the Department of Science and Technology,
Republic of the Philippines.
46
A critical tool that can be employed to address the problem [b.1.] Integration of local heuristic model with a stochastic
espoused in preceding discussions is the input-output (I-O) model – model for computing a probable global solution of the
an approach proposed by Wassily Leontief, which earned him the model.
Nobel Prize in Economic Sciences in 1973. The I-O model accounts
for sector interdependencies governing the magnitude and extent of [c] To investigate the effectiveness and sensitivity of the global
ripple effects due to changes in the economic structure of a region solution and develop policy insights for the Philippine government
or nation [8]. Recent extensions to I-O modeling have enhanced the through the following:
model's capabilities to account for the impact of an economic
perturbation, e.g., the inoperability I-O model [1], [9]. These models [c.1.] Testing the robustness of the global optimal solution
introduced sector inoperability, or the inability to satisfy as-planned with the changes in the budgetary and holding capacity
production levels, into I-O modeling [8]. Among these extensions, a parameters;
notable improvement was recently proposed by [1], which
incorporates inventory in a dynamic inoperability I-O model. Their
[c.2.] Comparing the nonlinear optimization model results
model, the Inventory Dynamic Inoperability I-O Model (I-DIIM),
with the conventional resource allocation approaches
has been illustrated in various applications, e.g., [3-4]. However,
using the data from Philippine I-O tables.
several aspects of the current formulation do not account for
complexities associated with inventory capacity and budget for
inventory reconfiguration. While inventory may promote resilience
against calamities, it is limited by economic sectors' capacity to keep I. METHODOLOGY
it and governments' ability to intervene in its configuration. For
brevity, the details of the I-O model and its recent extensions leading A. Model Development
to the I-DIIM are presented in Appendices A-D.
To address the gaps espoused in previous discussions, this
B. The Research Gaps work extends the I-DIIM through a Subsidy Optimization Model
(SOM) – a nonlinear programming model based on the I-DIIM that
An issue with the present formulation of the I-DIIM lies determines the optimal allocation of subsidy to minimize calamity-
in its inability to suggest initial inventory levels, minimize induced total economic loss. The formal definition of subsidy is as
inoperability, and, in turn, economic loss. This analysis is essential follows:
in the decision to appropriate resources for the reconfiguration of
initial inventory levels as a mitigation strategy to increase economic (0)
sectors' resilience. Galbusera et al. [3] proposed an optimization Definition 1. (Subsidy) Let 𝑠̃𝑖 denote the existing
model to supplement the need for such an analysis; however, their inventory levels of sector 𝑖 at 𝑡 = 0. The interest of this work lies in
(0)
model is impractical in several aspects. For instance, their analysis reconfiguring the size of these inventories through a subsidy, 𝑠̂𝑖 ,
is built on the assumption that economic sectors are willing to invest that comes from the government or some other financing body
in the reconfiguration of their inventories in preparation for a concerned with the macroeconomic consequences of a calamity.
calamity (e.g., pandemic). However, organizations hardly ever (0)
This idea builds on the assumption that 𝑠̂𝑖 is directly converted into
prepare for such calamities because they rarely occur. Unlike these
some sort of inventory that is readily convertible to revenue upon its
organizations, a central regulating body like the government is
allocation to sector 𝑖. In line with this, the initial inventory level
concerned with this problem. Government intervention to aid (0)
economic sectors is common during calamities because it is their 𝑠𝑖 of sector 𝑖 can be represented as the sum of its existing inventory
prime goal to maintain a stable national economy. It commonly does level and corresponding government subsidy as illustrated in (1).
so by appropriating limited financial subsidies to economic sectors.
While this is an obvious fact, no work has yet been made to (0) (0) (0)
𝑠𝑖 = 𝑠̃𝑖 + 𝑠̂𝑖 , ∀𝑖 ∈ ℕ (1)
empirically determine the optimal allocation of such subsidies to
disrupted economic sectors as a divisible resource. Assuming that (0) (0)
the financial subsidy is converted into some sort of inventory upon where 𝑠̃𝑖 , 𝑠̂𝑖 ∈ ℝ+ .
allocation to the sectors, the problem is now shifted on the allocation
of subsidy to configure existing inventory levels, reduce The summary of parameters, objective function,
inoperability, and minimize the overall economic consequence of a constraints, and the SOM formulation are presented in the following
calamity. In addressing the main problem espoused in the preceding sections.
discussions, the following objectives are outlined in this work:
Summary of Parameters
[a] To develop a nonlinear optimization model for the optimal
allocation of subsidy to configure the inventory levels of ℕ, set of economic sectors
disrupted, interdependent economic sectors through the following: 𝑛, number of economic sectors
ℎ, number of time-periods
[a.1.] Identification of parameters for the model that 𝛽𝑖 , inventory capacity of sector 𝑖
sufficiently reflect a real macroeconomic system; 𝛼, total subsidy budget
(0)
𝑠̃𝑖 , existing inventory level of sector 𝑖
[a.2.] Formulation of relations to distinguish budgetary (𝑡)
𝑠𝑖 , inventory level after subsidy of sector 𝑖 at time 𝑡
and holding capacity constraints while integrating them in (𝑡)
the model and integrating the I-DIIM in the model as a 𝑝𝑖 , production inoperability of sector 𝑖 at time 𝑡
(𝑡)
component of the objective function. 𝑥𝑖 , as-planned production of sector 𝑖 at time 𝑡
∗,(𝑡)
𝑐𝑖 , demand perturbation of sector 𝑖 at time 𝑡
[b] To develop an approach for solving the global solution of the 𝑘𝑖 , resilience coefficient of sector 𝑖
nonlinear optimization model through the following: ∗
𝑎𝑖𝑗 , interdependency coefficient of sector 𝑖
47
𝑡, time or period ∑𝑛𝑖=1 𝑠̂𝑖(0) ≤ 𝛼 (7)
(𝑡)
𝑞𝑖 , general inoperability of sector 𝑖 at time 𝑡
(0)
𝑠̂𝑖 , subsidy of sector 𝑖 The Subsidy Optimization Model
(0) (0)
Definition 16. Let vector 𝒘 denote the non-basic variables
𝑠̃𝑖 + 𝑠̂𝑖 ≥ 0, ∀𝑖 ∈ ℕ (5) and vector 𝒚 denote basic variables, which is a function of the basic
variables, i.e., 𝒚(𝒘). Then the GRG algorithm is illustrated as:
(0) (0)
𝑠̃𝑖 + 𝑠̂𝑖 ≤ 𝛽𝑖 , ∀𝑖 ∈ ℕ (6)
𝒁(𝒘) = 𝒛(𝒚(𝒘), 𝒘) (9)
Definition 15. (Total subsidy budget) Let 𝛼 denote the
total budget for the subsidy. This parameter characterizes the 𝑮(𝒘) = 𝒈(𝒚(𝒘), 𝒘) = 0 (10)
government's limited capacity to cushion the consequences of a
calamity through financial aid, e.g., subsidy, to its economic sectors. now,
This parameter is illustrated as follows:
48
𝑑𝑮 𝜕𝒈 𝜕𝒚 𝜕𝒈 Step 6: (Determine the inventory parameters) Determine
=0= + (11)
𝑑𝒘 𝜕𝒚 𝜕𝒘 𝜕𝒘 (0)
the existing inventory, 𝑠̃𝑖 , and inventory capacity, 𝛽𝑖 , for all
𝑑𝒁 𝜕𝒛 𝜕𝒛 𝜕𝒚 sectors.
= + (12)
𝑑𝒘 𝜕𝒘 𝜕𝒚 𝜕𝒘
Step 7: (Determine the subsidy parameters and set the
then from (11) and (12) the following is obtained: subsidy to zero for all sectors) Determine the subsidy budget, 𝛼.
(0)
Also, set the subsidy 𝑠̂𝑖 = 0 for all sectors and decide on the last
𝑑𝒁 𝜕𝒛 𝜕𝒛 𝜕𝒈 −1 𝜕𝒈 time-period, ℎ, of the simulation.
= − ( ) (13)
𝑑𝒘 𝜕𝒘 𝜕𝒚 𝜕𝒚 𝜕𝒘
Subject to: 𝑙𝒘 ≤ 𝒘 ≤ 𝑢𝒘 (14) Step 10: (Solve the SOM using the GRG with Muti-Start)
Solve (8) using the GRG with Multi-Start. In this routine, a set of
For brevity, solving the reduced problem of (14) is not 100 different feasible starting points are generated at first. The GRG
explained in this work. It is mostly based on the principles and logic algorithm is then applied to each of these feasible starting points turn
developed by [13] in conjunction with the Multi-Start search routine. by turn using (9)-(14), and the local optimal solutions for each
The GRG with Multi-Start is executed in this work using Visual starting point are stored. Finally, when all 100 different starting
Basic for Applications in Microsoft Excel. The execution of the points have been exploited, the solution that minimizes the objective
GRG with Multi-Start is detailed in the following discussions. function is the best optimal solution. This optimal solution,
∗,(0) ∗,(0) ∗,(0)
𝐸𝐿∗ (𝑠̂1 , 𝑠̂2 , … , 𝑠̂𝑛 ), is considered the closest to the global
C. The Subsidy Optimization Procedure solution.
Before solving the SOM, the I-DIIM is simulated for ℎ
time-periods. Also, because the model is solved multiple times using Steps 1-7 details the parameters' identification and
(0) calculation, while Steps 8-9 details the procedures for executing the
the GRG with Multi-start, 𝑠̂𝑖 = 0 is initially set for each solution.
I-DIIM. Lastly, Step 10 details the procedure for solving the SOM
The following subsidy optimization procedures were carried out to using the GRG with Multi-Start. To illustrate the proposed SOM's
solve the SOM: applicability, a case study in the Philippines during the COVID-19
pandemic is employed. The following sections detail the case study
Step 1: (Identify the system elements) Identify the background, datasets used in the case study, and the conventional
elements or sectors of the interdependent system. This system may allocation of subsidies versus the allocation of subsidies based on
be in any size and form, encompassing that of an interdependent the SOM results.
macroeconomic system.
D. Case Study Background: The Philippine Economy during the
Step 2: (Obtain the I-O matrix) Obtain the transactional COVID-19 Pandemic
data, 𝑥𝑖𝑗 , of the sectors organized as an I-O matrix. The data in the
I-O matrix must include total input, 𝑥𝑗 , total output, 𝑥𝑖 , and total Since the early stages of the COVID-19 pandemic, the
Philippine government has commenced measures to curb the spread
final demand, 𝑐𝑖 . Note that the as-planned final demand, 𝑐̂𝑖 , may be
of the virus, which include travel bans, various levels of community
estimated from the total final demand, 𝑐𝑖 , of the I-O matrix.
quarantines, and the T3 (test, trace, and treat) program [14], [15]. On
March 25, 2020, the Republic Act No. 11469 or the Bayanihan to
Step 3: (Determine the calamity impact parameters) Heal as One Act (BHOA) was enacted to implement these measures
Determine the calamity situation and its impact on the sectors, fully. A few months after its enactment, the Department of Health
∗,(𝑡)
including the degraded demand, 𝑐̃𝑖 , demand perturbation, 𝑐𝑖 , at (DOH) announced on May 05, 2020, in a press release that the
(0) country is starting to achieve the "flattening of the COVID-19
period 𝑡 and the initial production inoperability, 𝑝𝑖 .
curve" [16]. In the same report, the DOH warned on the possible
resurgence of cases if the people become complacent about social
Step 4: (Calculate the interdependency parameters)
distancing, mostly when the "enhanced community quarantine"
Calculate the Leontief technical coefficient of all sectors, 𝑎𝑖𝑗 ,
∗ (ECQ) – the strictest quarantine level in the country – is lifted [16].
interdependency coefficients, 𝑎𝑖𝑗 , using the procedures in As predicted, upon the easing of lockdown protocols on June 1,
Appendices A and B. 2020, the country had seen infections rising again, topping 2,000 in
recent days from about 500 when the lockdown was in effect [17].
Step 5: (Calculate the recovery and resilience The resurgence of cases disrupted government efforts to restart the
parameters) Calculate the recovery period, 𝑓𝑖 , and the production economy. It also prompted the government to place the country's
(𝑓 )
inoperability, 𝑝𝑖 𝑖 , at the recovery period, 𝑓𝑖 , for all sectors. Also, leading economic regions (e.g., Manila and Cebu) under strict
obtain the resilience coefficient, 𝑘𝑖 , for all sectors using the lockdown protocols multiple times from June to August 2020.
procedures in Appendices C and D. Despite the continuous surge of new COVID-19 cases, the country's
lockdown protocols are generally eased by the government since the
middle of August 2020, in an attempt to kick-start economic
recovery [18]. As of September 28, 2020, the country has 304,226
49
reported COVID-19 cases, ranking 8th in Asia and 21st worldwide in protocol in the Philippines. The datasets for 𝑨, 𝒙(𝑡) , and 𝒑(0) are as
terms of the highest reported cases [19]. follows:
50
60.0 600.0
non-allocation of subsidy to the large sectors is that the SOM has
exploited the economic sectors' independencies to prioritize the
50.0 500.0 small sectors. As a result of such allocation, the inoperability of the
40.0 400.0
small sectors is delayed and reduced overtime; because they
influence the large sectors, as a consequence, the inoperability of the
30.0 300.0 large sectors is reduced as well – leading to the minimization of the
20.0 200.0
total economic loss. Therefore, this work demonstrates that to
reduce total economic loss, prioritizing the large sectors in the
10.0 100.0
allocation of subsidy, like conventional methods, is not necessarily
0.0 0.0 an optimal strategy. In line with this, it is worthwhile to investigate
L1 L2 L3 S4 S5 S6 S7 S8 S9 S10 S11
Subsidy 0.0 0.0 1.3 50.1 33.9 32.2 31.0 8.2 19.0 17.0 4.3
when would such an unconventional subsidy allocation remains an
Economic Loss 569.6 189.9 76.9 41.2 30.7 34.4 27.8 38.4 11.5 12.6 0.0 optimal strategy. In line with this inquiry, the following section
investigates the sensitivity of the optimal solution with changes in
Subsidy Economic Loss
the subsidy budget, 𝛼, and inventory capacity, 𝛽𝑖 .
When it is economically disadvantageous to subsidize the Fig. 2. Subsidy of sectors (in BPhp) for different values of the subsidy budget
large sectors, those sectors that they are dependent on – the small (in BPhp)
sectors – may be subsidized instead. The small sectors provide
inputs required by the large sectors to stay operable. When the small For relatively smaller values of the subsidy budget, the
sectors become inoperable, it also affects the large sectors. While it large sectors are not prioritized. This observation is consistent with
is vital to reduce the large sectors' inoperability, reducing the small the observation in the preceding section. As the subsidy budget's
sectors' inoperability is also essential. However, when the subsidy value increases, the large sectors are prioritized except for L3
budget is small, prioritizing the large sectors would mean that the (Government Services). Note that L3 (Government Services) is
small sectors will become very inoperable, which would, in turn, among the least affected sectors during the COVID-19 pandemic,
result in the inoperability of the large sectors. A negative feedback (0)
with 𝑝11 = 0.13. This means that L3 (Government Services) is not
loop may occur in such an interdependent system, mainly because a priority for subsidy allocation because subsidizing it yields
the large sectors also provide inputs to the small sectors. Although minimal economic benefit, although it is among the large sectors.
the resilience coefficient, 𝑘𝑖 , of these sectors damps their This finding has several practical implications. For example, if the
inoperability overtime, leading to a recovery, analyzing their government decides to allocate a relatively small subsidy budget, it
interdependency to determine which sectors to prioritize in the needs to deal with political implications associated with the non-
allocation of subsidy is essential to speed up such recovery and prioritization of the large sectors. Such an unconventional approach
reduce total economic loss. However, the analysis of the complex to subsidy allocation might cause conflict between the large sectors
interdependency of the sectors in this respect is incredibly tricky. and the government.
Note that the SOM is developed within the framework of Large sectors are expected to be composed of large
the I-DIIM, which incorporates the interdependency of these sectors taxpaying institutions since their production volume is relatively
in the analysis of inoperability. Therefore, a logical reason for the
51
more extensive than in small sectors. Naturally, these sectors would The third case for sensitivity analysis intends to provide a
expect subsidy proportionate to their contribution to the national general perspective on the minimized total economic loss's
economy. The non-prioritization of large sectors can negatively sensitivity with simultaneous changes in both subsidy budget and
impact their confidence towards the government's policies and find inventory capacity values. A heat map in Fig. 4 presents the results
the country not secure for business. While the decision to least of this analysis. It can be vividly observed from the heat map that,
prioritize the large sectors in case of a small subsidy budget is for some value of the inventory capacity multiplier, there is a value
optimal, relaying the reason behind such a decision might be of the subsidy budget wherein the total economic loss can no longer
difficult. Such misunderstanding and conflict might result in large be improved. This value is indicated as the threshold line. The
businesses in these large sectors to transfer operations to other minimized total economic loss cannot be improved further than the
countries, which is terrible for the national economy. Thus, to avoid threshold subsidy that the economic sectors can utilize to
such a conflict its hefty consequences, the government may opt to reconfigure their initial inventory level due to their limited capacity
sufficiently increase its subsidy budget. For instance, when the to store inventory. Inventory capacity, therefore, limits the
subsidy budget is increased to 450.0 BPhp, the large sectors' utilization of subsidies, which can result in a surplus in the budget.
prioritization is an optimal strategy (see Fig. 2).
A surplus in the budget may be undesirable for several
Case 2: Changing the Inventory Capacity reasons. For instance, it indicates inadequate planning resulting in
inefficient utilization of the subsidy budget. For a government, the
The second case for sensitivity analysis is as follows: surplus money could have been invested or used for other profound
Suppose the Philippine government intends to impose a policy that means of public spending to serve the greater good. This finding has
requires the sectors to have a minimum inventory capacity as a several practical implications. For example, in the Philippines, this
disaster preparedness strategy. To determine the appropriate finding is of great importance, especially with the present debate on
inventory capacity to require from the sectors, it is essential to the suitability of the budget allotted for its proposed economic
determine how the optimal solution changes with the values of stimulus bill (House Bill 6815) in response to the COVID-19
parameter 𝜔 – the inventory capacity multiplier. In this regard, the pandemic. This study's findings may be utilized to adjust the debated
SOM is solved for multiple values of the inventory capacity budget to meet the government's criteria without compromising the
multiplier at a constant subsidy budget 𝛼 = 197.0 BPhp. The results efficiency of utilizing the subsidy budget. In general, this finding
of this analysis are presented in Fig. 3. suggests that, as much as possible, a decision-making body, e.g., a
government, must ensure that, for a given capacity of economic
1
sectors to keep inventory, an appropriate budget is decided such that
4.95 5 1.051.1 1.15
4.85 4.9 80 1.2 surplus is avoided.
4.8 1.25
4.75 1.3
4.7 70 1.35
4.65 1.4
4.6 1.45
4.55 60 1.5
L1 4.5 1.55 Subsidy Budget, 𝛼 (in BPhp)
4.45 50 1.6
L2 4.4 1.65
4.35 40 1.7
L3 4.3
30
1.75
4.25 1.8
S4 4.2 20 1.85
4.15 1.9
S5 4.1 10 1.95
S6 4.05 0 2
4 2.05
S7 3.95 2.1
3.9 2.15
S8 3.85 2.2
3.8 2.25
S9
3.75 2.3
S10 3.7 2.35
3.65 2.4
S11 3.6 2.45
3.55 2.5
3.5 2.55
3.45 2.6
Inventory Capacity Multiplier, 𝜔
3.4 2.65
3.35 2.7
3.3 2.75
3.25 3.2 2.852.8
3.15 3.13.05 3 2.952.9
Fig. 3. Subsidy of sectors (in BPhp) for different values of the inventory
capacity multiplier
52
C. Comparative Case Analysis Scenario 1 Scenario 2 Scenario 3
59.0
demonstrate the importance of an inventory subsidy during
50.1
calamities, (2) show how decisions derived from the SOM can
deviate from the conventional method of appropriating resources,
33.9
32.2
31.0
and (3) contrast the results of the two methodologies. There are three
28.8
26.9
scenarios considered in this analysis, which are elaborated as
19.5
19.0
17.0
13.2
12.6
12.1
follows:
9.3
8.2
7.4
6.6
4.3
1.7
1.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Scenario 1: No Subsidy is Allocated
L1 L2 L3 S4 S5 S6 S7 S8 S9 S10 S11
11.65%
8.58%
8.52%
7.91%
7.66%
7.61%
7.45%
6.97%
4.97%
4.57%
4.50%
3.97%
3.30%
3.30%
2.41%
2.28%
2.03%
0.09%
0.06%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
L1 L2 L3 S4 S5 S6 S7 S8 S9 S10 S11
53
merely assumes that the economic sectors are independent system
L1 L2 L3
elements. This assumption is the main reason for its wide adoption
1.00 0.80 0.25 among analysts and decision-makers in appropriating divisible
Inoperability, 𝒒
0.70
0.80 0.60 0.20 resources (e.g., subsidy), hence being conventional. While it is
Inoperability, 𝒒
Inoperability, 𝒒
0.60 0.50 0.15
0.40
simple, it doesn't mean that it is optimal. Such analysis fails to
0.40 0.30 0.10
0.20
account for and take advantage of the complex interdependencies of
0.20 0.05
0.10 the economic sectors. These interdependencies are incredibly
0.00 0.00 0.00
0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30 challenging to perceive or even infer from. Thus, exploiting them is
Period, t (in months) Period, t (in months) Period, t (in months) difficult for analysts or decision-makers. However, this work
demonstrates the significance of such an analysis in reducing
calamity-induced inoperability (see Fig. 8), leading to the
S4 S5 S6 minimization of total economic loss consequent to it. Hence, the
0.70 0.80 1.00
SOM's importance is its ability to guide analysts and decision-
0.60 0.70 makers to make systemic and systematic resource allocation
0.60 0.80
Inoperability, 𝒒
Inoperability, 𝒒
Inoperability, 𝒒
0.50
0.40
0.50 0.60 strategies.
0.40
0.30 0.40
0.30
0.20 0.20 0.20
0.10
0.00
0.10
0.00 0.00
IV. CONCLUSION
0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30
Period, t (in months) Period, t (in months) Period, t (in months) There have been several extensions to the classical
Leontief Input-Output model applied in infrastructure risk research.
Among the said extensions is the Inventory Dynamic Inoperability
S7 S8 S9
Input-Output Model (I-DIIM). The I-DIIM illustrated the
0.70 0.80 0.70 importance of inventory to the preparedness and resilience of
0.60 0.70 0.60
0.60
infrastructures in a production environment, e.g., a macroeconomic
Inoperability, 𝒒
Inoperability, 𝒒
Inoperability, 𝒒
0.50 0.50
0.50 system, to disruptive events. Although the economic implications of
0.40 0.40
0.40
0.30 0.30
0.20
0.30
0.20
a disruptive event to a macroeconomic system are primarily the
0.20
0.10 0.10 0.10 government's concern, no work within the domains of infrastructure
0.00 0.00 0.00
0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30
risk research has yet attempted to provide a way to empirically
Period, t (in months) Period, t (in months) Period, t (in months) determine the optimal intervention required from governments to
facilitate the preparedness and resilience of economic sectors. This
work addressed this major gap in the literature by developing a
S10 S11 Subsidy Optimization Model (SOM) that is based on the I-DIIM
0.01 with the following key model characteristics: (1) distinguishes
0.80
0.70 0.01 inventory constraints from budgetary constraints, (2) incorporates
Inoperability, 𝒒
0.60 0.01
Inoperability, 𝒒
0.50 Scenario 1
subsidy as a means to configure initial inventories, and (3) integrates
0.01
0.40 the I-DIIM in a nonlinear optimization model for minimizing total
0.00 Scenario 2
0.30
Scenario 3
0.20 0.00 economic loss within an exogenously defined simulation period.
0.10
0.00 0.00 This work also proposes a simulation-based subsidy optimization
0 5 10 15 20 25 30
0 5 10 15 20 25 30
Period, t (in months)
procedure with the Generalized Reduced Gradient (GRG) with the
Period, t (in months)
Multi-Start algorithm for solving the SOM. The SOM's applicability
is tested through an illustrative case study in the Philippines during
the COVID-19 pandemic.
Fig. 8. I-DIIM simulation results on the inoperability of the economic
sectors in Scenarios 1, 2, and 3
This work revealed that the SOM differs from
conventional methods for allocating divisible resources at low
Fig. 6 presents the economic loss in each scenario over subsidy budgets. For instance, at a subsidy budget of 197.0 BPhp,
time. Considering Scenario 1 (No subsidy is allocated) as the the SOM's optimal solution indicated that large sectors be least
baseline case (the worst case), Scenario 2 (Conventional subsidy prioritized in the allocation of subsidy. Such peculiar results were
allocation) reduced the total economic loss only by 21.6%, while found consequent to the interdependency of the economic sectors,
Scenario 3 (Optimal subsidy allocation) reduced the total economic which the SOM was able to exploit to minimize the total economic
loss by 33.8%. The contribution of each sector in each scenario in loss. The robustness of such findings was tested through sensitivity
reducing total economic loss is presented in Fig. 7. In Scenario 3 analysis by changing the subsidy budget and inventory capacity
(Optimal subsidy allocation), although little to no subsidy was values. Consistent with initial results, it was found that the large
allocated for the large sectors, they were among the most significant sectors are least prioritized for relatively smaller values of the
contributors in reducing the total economic loss. While this subsidy budget; however, as the subsidy budget's value increases,
observation on the contribution of the large sectors towards the the priority shifts towards the large sectors except for L3
reduction of total economic loss is also true in Scenario 2 (Government Services), which is among the least affected sectors
(Conventional subsidy allocation), it is worth noting that, for the rest during the COVID-19 pandemic in the Philippines. This result
of the sectors, Scenario 3 (Optimal subsidy allocation) far somehow resembles the conventional method for the allocation of
outperforms it in terms of total economic loss reduction. These divisible resources. It was also found that when the inventory
results vividly illustrate the strength of Scenario 3 (Optimal subsidy capacity is low, the large sectors are prioritized in the allocation of
allocation) over Scenario 2 (Conventional subsidy allocation), subsidy. However, as the inventory capacity increases, the priority
which is primarily grounded on the ability of the SOM to exploit the diverges from the large sectors towards the small sectors, which is
interdependency of the sectors. still instigated by the complex interdependency of the sectors
exploited by the SOM.
The strength of Scenario 2 (Conventional subsidy
allocation) is the simplicity of its analytical procedures because it
54
Furthermore, this work investigated how simultaneous [4] E. M. Y. Jabilles, J. M. T. Cuizon, P. M. A. Tapales, R. L.
changes in the subsidy budget and inventory affects the Urbano, L. A. Ocampo, and D. A. A. Kilongkilong,
minimization of total economic loss. In this work, the analysis found "Simulating the impact of inventory on supply chain resilience
out that there is a corresponding value of the subsidy budget for with an algorithmic process based on the supply-side dynamic
some inventory capacity multiplier values wherein the total inoperability input–output model," International Journal of
economic loss can no longer be improved. The finding is essential Management Science and Engineering Management, vol. 14,
for avoiding utilization inefficiencies and budget surplus, which is no. 4, pp. 253–263, Oct. 2019, doi:
essential for decision-makers and policymakers. Specifically, the 10.1080/17509653.2018.1555693.
finding suggests that, as much as possible, a decision-making body, [5] L. Xu, Q. Zhang, K. Wang, and X. Shi, "Subsidies, loans, and
e.g., a government, must ensure that, for a given capacity of companies' performance: evidence from China's photovoltaic
economic sectors to keep inventory, an appropriate budget is industry," Applied Energy, vol. 260, p. 114280, Feb. 2020, doi:
decided such that surplus is avoided. 10.1016/j.apenergy.2019.114280.
[6] C. M. Himang et al., "Acceptability evaluation of a developed
Through comparative case analysis, this work contrasted thermal infrared device for fire risk management using the
conventional methods of subsidy allocation, e.g., proportional unified theory of acceptance and use of technology (UTAUT)
allocation, and the allocation scheme proposed using the SOM. The model," International Journal of Sociotechnology and
analysis vividly demonstrated the inadequacy of treating economic Knowledge Development, vol. 12, no. 3, pp. 1–25, Jul. 2020,
sectors in isolation, which is employed by conventional allocation doi: 10.4018/IJSKD.2020070101.
methods. The analysis also highlighted the SOM's ability to exploit [7] E. F. Selerio Jr., R. D. Arcadio, G. J. Medio, M. A. Nacua, and
and leverage economic sectors' interdependency to develop an G. A. Pedregosa, "An Application of a Hybrid Fuzzy Multi-
allocation scheme that minimizes total economic loss, which would Criteria Decision-Making Approach in Managing Equipment
otherwise be an incredibly difficult task to fulfill without it. The Hazards in Woodworking," International Journal of Safety and
SOM, therefore, is instrumental in guiding analysts and decision- Security Engineering, vol. 10, no. 4, pp. 517–527, Aug. 2020,
makers in making systemic and systematic resource allocation doi: 10.18280/ijsse.100411.
strategies. [8] J. R. Santos, M. J. Orsi, and E. J. Bond, "Pandemic Recovery
Analysis Using the Dynamic Inoperability Input-Output
For future work, researchers may explore a SOM Model," Risk Analysis, vol. 29, no. 12, pp. 1743–1758, Dec.
modification to consider allocating subsidies in multiple schedules. 2009, doi: 10.1111/j.1539-6924.2009.01328.x.
In the real world, large budgets for subsidies are seldom allocated at [9] J. R. Santos and Y. Y. Haimes, "Modeling the demand
once. Instead, it is allocated in tranches. Scheduling each tranche reduction input-output (I-O) inoperability due to terrorism of
such that total economic loss is minimized within a simulation interconnected infrastructures," Risk Analysis, vol. 24, no. 6,
period may be a worthwhile undertaking for a SOM extension. One pp. 1437–1451, Dec. 2004, doi: 10.1111/j.0272-
of the major limitations of this work and preceding works on I-DIIM 4332.2004.00540.x.
is the assumption that inventories are readily convertible to revenue. [10] R. Sharma and B. Glemmestad, "On generalized reduced
In this regard, researchers may also consider developing new gradient method with multi-start and self-optimizing control
constraints to consider the processing capacity and processing time structure for gas lift allocation optimization," Journal of
of inventories to address this limitation. The concepts mentioned Process Control, vol. 23, no. 8, pp. 1129–1140, Sep. 2013, doi:
above are not addressed in the present work because it intends to 10.1016/j.jprocont.2013.07.001.
serve as a groundwork for exploring such concepts in the following [11] P. Wolfe, "Methods of Nonlinear Programming, in: Graves,
series of publications on this topic. R.L. & Wolfe, P. (Eds.), Recent Advances in Mathematical
Programming, McGraw-Hill, New York, 67-86.," 1963.
[12] J. Abadie and J. Carpentie, "Generalization of the Wolfe
reduced gradient method to the case of nonlinear constraints,
ACKNOWLEDGEMENT
in: Fletcher, R. (Eds.), Optimization, Academic Press, New
York, USA, 37–47.," 1969.
The author expresses appreciation to Engineering
[13] L. S. Lasdon, R. L. Fox, and M. W. Ratner, “Nonlinear
Research and Development for Technology (ERDT) program of the
Optimization Using the Generalized Reduced Gradient
Department of Science and Technology, Republic of the Philippines
Method. Revue française d’automatique, informatique,
who funded this work, as part of the principal author’s (EFSJ)
recherche opérationnelle,” RAIRO - Operations Research -
master’s thesis, through the Thesis and Dissertation Grant (TDG).
Recherche Opérationnelle, vol. 8, no. V3, pp. 73–103, 1974,
Accessed: Sep. 17, 2020. [Online]. Available:
REFERENCES http://www.numdam.org/item?id=RO_1974__8_3_73_0.
[1] K. Barker and J. R. Santos, "Measuring the efficacy of [14] M. M. Ong, R. M. Ong, G. K. Reyes, and L. B. Sumpaico-
inventory with a dynamic input–output model," International Tanchanco, "Addressing the COVID-19 Nutrition Crisis in
Journal of Production Economics, vol. 126, no. 1, pp. 130–143, Vulnerable Communities: Applying a Primary Care
Jul. 2010, doi: 10.1016/J.IJPE.2009.08.011. Perspective," Journal of Primary Care and Community Health,
[2] O. Jonkeren and G. Giannopoulos, "Analysing Critical vol. 11. SAGE Publications Inc., 2020, doi:
Infrastructure Failure with a Resilience Inoperability Input– 10.1177/2150132720946951.
output Model," Economic Systems Research, vol. 26, no. 1, pp. [15] G. D. Legaspi et al., "Letter to the Editor' Service and Training
39–59, Jan. 2014, doi: 10.1080/09535314.2013.872604. During the COVID-19 Pandemic: Perspectives from a
[3] L. Galbusera, I. Azzini, O. Jonkeren, and G. Giannopoulos, Neurosurgical Center in the Philippines,'" World
"Inoperability Input-Output Modeling: Inventory Optimization Neurosurgery, vol. 139. Elsevier Inc., pp. 741–743, Jul. 01,
and Resilience Estimation during Critical Events," ASCE- 2020, doi: 10.1016/j.wneu.2020.05.138.
ASME Journal of Risk and Uncertainty in Engineering [16] DOH, "PH Starts to Flatten Curve; ECQ, Physical Distancing
Systems, Part A: Civil Engineering, vol. 2, no. 3, p. B4016001, Pay Off," Department of Health, 2020.
Sep. 2016, doi: 10.1061/AJRUA6.0000861.
55
[17] The Straits Times, "Coronavirus: Resurgence of cases threatens where
bids to kick-start Philippine economy," The Straits Times,
2020. 𝑥𝑖𝑗
[18] CNN Philippines, "Metro Manila, 4 provinces to ease into GCQ 𝑨 = [𝑎𝑖𝑗 ] =[ ] , (16)
𝑛×𝑛 𝑥𝑗
𝑛×𝑛
starting Aug. 19," CNN Philippines , 2020.
[19] ECDC, "COVID-19 situation update worldwide, as of 28 where 𝑥𝑖 is the total production output of sector 𝑖 while the
September 2020," European Centre for Disease Prevention and Leontief technical coefficient 𝑎𝑖𝑗 is the ratio of the output of sector
Control, 2020.
𝑖 that serves as an input for sector 𝑗 concerning the total production
[20] NEDA, "Statement of NEDA Acting Secretary Karl Kendrink
requirement of sector 𝑗; and 𝑐𝑖 is the final demand of the 𝑖th sector,
T. Chua On The 2020 Q2 Performance of the Philippine
Economy," National Economic and Development Authority, for 𝑖, 𝑗 ∈ ℕ.
2020.
[21] PSA, "Press Release," Philippine Statistics Authority, 2020. Appendix B. The Demand-driven Inoperability I-O Model (IIM)
[22] TESDA, "COVID-19's Impact to the Philippine
EconomSectors: A TVET Response. TVET Brief: Industry While the Classical I-O Model has gained prominence in
Trends," Technical Education and Skills Development its ability to integrate economic sectors' interdependence, its
Authority, 2020. applications are quite limited due to its inability to assess economic
[23] PNA and F. M. Cervantes, "House okays ARISE bill on 3rd risks explicitly. Based on (15) and (16), the IIM was proposed by [9]
reading | Philippine News Agency," 2020. to address such a gap. They employed the IIM in assessing the
[24] PSA, "Input Output Tables," Philippine Statistics Authority, impact of a disruption and its spread across an economic network.
2017. The capability of the IIM to model the propagation of inoperability
[25] ADB, "ADB Philippine Enterprise Survey on COVID-19 in a system makes it a relevant framework in the development of
Impact | ADB Data Library," Asian Development Bank, 2020. risk-management policies [9]. The following definitions constitute
https://data.adb.org/dataset/adb-philippine-enterprise-survey- the IIM:
covid-19-impact (accessed Sep. 18, 2020).
[26] ADB, "Philippine Economy to Decline Further in 2020 Amid Definition 2. (Demand-side perturbation) Let 𝒙 be the as-
COVID-19, With Recovery in 2021," Asian Development planned production vector that describes the as-planned production
Bank, 2020. output of each sector. Provided that each sector is defined on the
[27] W. W. Leontief, "Quantitative Input and Output Relations in premise of "economic significance" [9], then [𝑑𝑖𝑎𝑔(𝒙)]−1 always
the Economic Systems of the United States," The Review of exist since it guarantees that 𝑥𝑖 ≠ 0, for 𝑖 ∈ ℕ. The preceding
Economics and Statistics, vol. 18, no. 3, p. 105, Aug. 1936, doi: argument also suggests that 𝑑𝑖𝑎𝑔(𝒙) is always a nonsingular matrix,
10.2307/1927837. which sufficiently establishes its inverse's existence. Denoting 𝒄̂ as
[28] Y. Y. Haimes, B. M. Horowitz, J. H. Lambert, J. R. Santos, C. the as-planned final demand and 𝒄̃ as the degraded demand which
Lian, and K. G. Crowther, "Inoperability Input-Output Model may be brought about by fear and unrest due to some disaster, the
for Interdependent Infrastructure Sectors. I: Theory and demand-side perturbation vector 𝒄∗ is defined as the decrease in the
Methodology," Journal of Infrastructure Systems, vol. 11, no. final demand (𝒄̂ − 𝒄̃), normalized according to the as-planned
2, pp. 67–79, Jun. 2005, doi: 10.1061/(ASCE)1076- production 𝒙, which is illustrated as follows:
0342(2005)11:2(67).
[29] R. , Miller and P. Blair, Input output analysis foundations and 1
0 ⋯ ⋯ 0
extensions, 2nd ed. Cambridge University Press, 2009. 𝑐1∗ 𝑥1 𝑐̂1 − 𝑐̃1
⋮ 0 ⋱ ⋱ ⋮ ⋮
1
𝒄∗ = [𝑐𝑖∗ ]𝑛×1 = [𝑑𝑖𝑎𝑔(𝒙)]−1 [𝒄̂ − 𝒄̃] ⇔ 𝑐𝑖∗ = ⋮ ⋱ 𝑥𝑖
⋱ ⋮ 𝑐̂𝑖 − 𝑐̃𝑖 (17)
Appendix A. The Classical I-O Model ⋮ ⋮
⋮ ⋱ ⋱ 0
[𝑐𝑛∗ ] 1 [𝑐̂𝑛 − 𝑐̃𝑛 ]
[0 ⋯ ⋯ 0 𝑥𝑛 ]
The I-O analysis is a form of macroeconomic analysis
based on the interdependencies between economic sectors or The 𝑖th element in 𝒄∗ represents the demand-side
industries, showing how the output from one industrial sector may perturbation of the 𝑖th sector. Denoting this scalar element by 𝑐𝑖∗ , it
become an input to another industrial sector. This method is can be extracted from (17) as follows:
commonly employed for determining the impacts of economic
disruptions and their ripple effects throughout an economy [27]. The 𝑐̂𝑖 −𝑐̃𝑖
𝑐𝑖∗ = , (18)
column entries of the intersectoral matrix employed in I-O models 𝑥𝑖
represent the input to an economic sector, while the row entries
represent outputs from a given sector. The intersectoral matrix where 𝑐̂𝑖 is the as-planned demand of sector 𝑖, 𝑐̃𝑖 is the
presents each sector's dependence on every other sector, both as a 𝑐̂
corresponding degraded demand of sector 𝑖, 𝑥𝑖 ∈ ℝ+ , 𝑐𝑖∗ ∈ [0, 𝑖 ],
customer of outputs from other sectors and as a supplier of inputs. 𝑥𝑖
Each column of the matrix indicates the monetary value of inputs to and 𝑖 ∈ ℕ.
each sector, while each row represents the value of each sector's
outputs. Building on this idea, the Classical I-O Model is defined as Definition 3. (Interdependency matrix) Let 𝑨∗ be the
follows: demand-side interdependency matrix that characterizes the degree
of coupling of the sectors. The relationship of matrix 𝑨∗ to the
Definition 1. (Classical I-O model) Let ℕ = {1,2, … , 𝑛} Leontief technical coefficient matrix 𝑨 and the as-planned
be a set of economic sectors. Assuming that each sector sells some production vector 𝒙 is illustrated as follows:
of its output to other sectors (intermediate output) and some of its
output to consumers (final output, or final demand), the Classical I-
O Model [27] is expressed as follows:
56
∗
𝑨∗ = [𝑎𝑖𝑗 ] = [[𝑑𝑖𝑎𝑔(𝒙)]−1[𝑨][𝑑𝑖𝑎𝑔(𝒙)]]𝑛×𝑛 ⇔ transforming the resulting equation into the normalized inoperability
𝑛×𝑛
1
0 ⋯ ⋯ 0 𝑎11 ⋯ 𝑎1𝑗 ⋯ 𝑎1𝑛 𝑥1 form (as in Definition 2-4) yields the discrete form of the DIIM that
𝑥1 0 ⋯ ⋯ 0
0 ⋱ ⋱ ⋮ ⋮ ⋱ ⋮ ⋮ 0 ⋱ ⋱ ⋮ presents the inoperability vector 𝒒(𝑡+1) , as illustrated in (24).
1
⋮ ⋱ ⋱ ⋮ 𝑎𝑖1 ⋯ 𝑎𝑖𝑗 ⋯ 𝑎𝑖𝑛 ⋮ ⋱ 𝑥𝑗 ⋱ ⋮ (19)
𝑥𝑖
⋮ ⋮ ⋱ 0 ⋮ ⋱ ⋱ 0 1
⋮ ⋱ ⋱ 0
1 [𝑎𝑛1 ⋯ 𝑎𝑛𝑗 ⋯ 𝑎𝑛𝑛 ] [ 0 ⋯ ⋯ 0 𝑥𝑛 ]
𝑲 = −𝑩−1 ⇔ 𝑘𝑖 = ; ∀𝑖 ∈ ℕ (23)
𝑏𝑖
[0 ⋯ ⋯ 0
𝑥𝑛 ]
(𝑡+1)
The derivation of the scalar representation of (19) is 𝒒(𝑡+1) = [𝑞𝑖 ] = 𝒒(𝑡) + 𝑲[𝑨∗ 𝒒(𝑡) + 𝒄∗,(𝒕) − 𝒒(𝑡) ], (24)
𝑛×1
omitted for brevity, but it can be found in [28]. The scalar
representation of (19) is illustrated as follows: where 𝑨∗ denotes the demand-side interdependency
(𝑡)
𝑥𝑗
matrix as defined in Definition 3, 𝑞𝑖 ∈ [0,1] ∀𝑖, 𝑡; and 𝑖 ∈ ℕ.
∗
𝑎𝑖𝑗 = 𝑎𝑖𝑗 ( ); for 𝑖, 𝑗 ∈ ℕ (20)
𝑥𝑖
Appendix D. The Inventory Dynamic Inoperability I-O Model
Definition 4. (Inoperability I-O model) Let 𝒒 be the final (I-DIIM)
inoperability vector that describes the inoperability of the different
sectors. Within the framework of the IIM, inoperability is defined as Barker and Santos [1] introduced the I-DIIM, which can
a level of a system's dysfunction expressed as a percentage of its as- model the effects of the resilience measures that delay the onset of
planned production capacity [9], [28]. Building on previous infrastructure or economic sectors' inoperability due to access to
definitions, 𝒒 is calculated as follows: some inventory. The definitions that constitute the I-DIIM are
presented as follows:
𝒒 = [𝑞𝑖 ]𝑛×1 = [𝑰 − 𝑨∗ ]−1 𝒄∗ , (21)
Definition 8. (Production inoperability) Let 𝒑(𝑡) =
(𝑡) (𝑡)
where 𝑰 is an identity matrix, 𝑞𝑖 ∈ [0,1], and 𝑖 ∈ ℕ. [𝑝𝑖 ] be the production inoperability vector, where 𝑝𝑖 denotes
𝑛×1
the production inoperability consequent only to physical disruption
The rigorous derivation of (21) is not presented herewith of processes of sector 𝑖 at the end of time 𝑡 [1]. The calculation of
for brevity, but it can be found in [9]. (𝑡)
𝑝𝑖 is illustrated as a decaying function of time 𝑡 as follows:
Appendix C. The Demand-driven Dynamic Inoperability I-O (𝑡) (0)
Model (DIIM) 𝑝𝑖 = 𝑒 −𝑘𝑖𝑡 𝑝𝑖 , (25)
(0)
The main limitation of the IIM is its inability to model the where 𝑝𝑖 is the exogenously defined initial production
concept of "dynamic economic resilience" [28]. The need to (𝑡)
inoperability of sector 𝑖, 𝑝𝑖 ∈ [0,1] ∀𝑖, 𝑡; 𝑖 ∈ ℕ, and 𝑘𝑖 is as
integrate the temporal dynamic behavior of industry recoveries in
defined in Definition 5 and is calculated using (26).
the analysis of inoperability leads to the development of the DIIM,
as an extension that supplements and complements the static IIM. In (0)
𝑝
the derivation of the DIIM, [28] introduced the concept of an ln[ 𝑖
(𝑓𝑖) ]
𝑝
industry resilience coefficient that is defined as follows: 𝑘𝑖 = 𝑖
, (26)
𝑓𝑖
57
𝒒(𝑡+1) = [𝑞𝑖
(𝑡+1)
] = 𝑞𝑖 (𝑠𝑖
(𝑡)
, 𝑝(𝑖𝑡) , 𝑥(𝑖𝑡) , 𝑐∗,(𝑡)
𝑖 , 𝑘𝑖 , 𝑎∗𝑖𝑗 ) =
𝑛×1
(𝑡) ∗,(𝑡) (𝑡+1) (𝑡+1) (𝑡+1)
𝐹𝑖 [𝒒 , 𝑐𝑖 ], 𝑖𝑓 𝑠𝑖 ≥ 𝑝𝑖 𝑥𝑖
(𝑡+1)
(𝑡+1) 𝑠𝑖 ∗,(𝑡) (𝑡+1) (𝑡+1) (𝑡+1)
max {[𝑝𝑖 − (𝑡+1) ] , 𝐹𝑖 [𝒒(𝑡), 𝑐𝑖 ]} , 𝑖𝑓 𝑠𝑖 ∈ (0, 𝑝𝑖 𝑥𝑖 )
𝑥𝑖
, (28)
(𝑡+1) ∗,(𝑡) (𝑡) (𝑡+1)
max {𝑝𝑖 , 𝐹𝑖 [𝒒(𝑡) , 𝑐𝑖 ]} , 𝑖𝑓 𝑠𝑖 > 0, 𝑠𝑖 =0
∗,(𝑡) (𝑡+1) (𝑡)
{ 𝐹𝑖 [𝒒(𝑡) , 𝑐𝑖 ]], 𝑖𝑓 𝑠𝑖 = 𝑠𝑖 =0
(𝑡)
where 𝑞𝑖 ∈ [0,1] ∀𝑖, 𝑡, and 𝑖 ∈ ℕ and at time 𝑡 = 0, the
(0)
initial inoperability 𝑞𝑖 varies with initial inventory scenarios as
follows:
58
Earthquake Resistant Building Through Genetic
Algorithm Optimization: A Case Study Using
Ground Motion Data
Lisa Guieb1, a *, Sherwin Guirnaldo², Glenn Lorenzoᵌ
¹Department of Civil Engineering, MSU – Iligan Institute of Technology, Iligan City;
²Department of Mechanical Engineering, MSU – Iligan Institute of Technology, 9200 Iligan City;
³Department of Civil Engineering, MSU – Iligan Institute of Technology, Iligan City;
a
lisatguieb@gmail.com
Abstract— The study uses the dynamic analysis design is necessary. It is directed to find optimal
procedure for Response Spectrum Analysis (RSA) for structural design resulting in size optimization,
buildings from the Site-specific Response Spectrum Curve minimum construction weight results in sustainable
obtained on October 2019, series of earthquakes that caused development with lesser damage to buildings. The
major damage to several structures in Mindanao.
Optimization techniques are used to solve structural
optimization of building design under seismic
engineering designs considering safety, stability, and excitation is crucial for optimal design parameters
sustainability when subjected to earthquake loadings. The among considered alternatives while ensuring design
STAAD.Pro structural analysis software has been used for safety. It provides the satisfaction of ultimate strength
the design computation. For this study, the member size of and serviceability limit states prescribed by design
columns and beams was checked for displacement, and codes.
forces considered as constraints for optimizing different
cross-sections using a genetic algorithm. The design with the 2. OPTIMIZATION PROBLEM
highest performance, which satisfies the code parameters,
are considered the best performing design that was chosen 2.1 Fundamental optimization problem
from the final generation out of one hundred population. The
numerical algorithm is developed in MATLAB software The objective function of the optimization is to
incorporating the response spectrum analysis of five-story minimize the sizes of beams and columns to minimize
story building and optimization routine. The most common
parameter used in this study is the minimization of inter-
the weight of structure and the displacement of the
story drift and dynamic base shear, which is used in seismic building under seismic actions. It is design to resist
design codes of the National Structural Code of the lateral forces such as earthquake forces which may
Philippines (NSCP). cause failure to structures.
Keywords: Response spectrum analysis; site-specific ground In this study, the design variables are associated
motion; response spectrum curve; optimization; genetic with dimensions, beam width, beam depth and column
algorithm member sizes. In the sizing optimization, ‘x’ is some
type of thickness i.e. cross sectional area of beam and
I. INTRODUCTION
column. The objective function is expressed in
On October 2019, series of earthquakes struck Equation 1.
Tulunan, Cotabato, Mindanao and vicinity. The 𝒇(𝑿) = 𝒇([𝑪𝒐𝒍𝒖𝒎𝒏 𝒔𝒊𝒛𝒆𝒔, 𝑩𝒆𝒂𝒎 𝒔𝒊𝒛𝒆𝒔])
Philippine Institute of Volcanology & Seismology = 𝐦𝐢𝐧(∆𝒙𝒅𝒓𝒊𝒇𝒕 )
(PHIVOLCS), reported that these earthquakes 𝒏
59
[𝒙𝒌 ]the displacement in the node k, p the total of (a)
node, 𝒙𝒕 the total drift, W the total weight of beam.
PLAN
b=19m
x1 x2 a1 =8m
a2=7m a=23m
x1 x3
a3=8m
ELEVATION
Roof
3.5m
Floor 5
3.5m
Floor 4
3.5m h
Floor 3
3.5m
Floor 2
4m
Ground
6m 7m 6m
b
concrete building.
Figure 3. Model 5-Storey RC
60
Genetic Algorithm (GA) is inspired by the process
of natural evolution (Goldberg 1989). In this
optimization approach, a population of individuals
5th Floor (variables) is evolved to produce best performance of
designs. The evolution process in this article is one
4th Floor
hundred generations continue. The evolution
3rd Floor progresses in generations. In each generation, the
fitness of individuals are selected randomly modified
2nd Floor
to produce the next generation; and continue until a
1st Floor maximum number of generations are formed or a
satisfactory level of fitness is achieved.
MATLAB
Figure 4. Lumped mass model with mass and stiffness Design Variables
More specifically, it aimed to understand the
structural behavior of buildings in Mindanao. The DESIGN CANDIDATES
purpose of this effort is to help reduce the impact of Constraint
future large quakes in the country particularly in the s
seismically active areas in the Philippines. It is an
advantage for engineers to understand the recent Fitness value
earthquake effects related to geology, ground motion, EARTHQUAKE or
earthquake intensity, and magnitude to address the SIMULATOR Performance
advancement in developing building codes and value
(STAAD Pro
designs that require site-specific analysis in the software)
Philippines.
EQ
II. METHODOLOGY Records
2.1 Structural Optimization
Figure 5. – Procedure of Design Optimization
Structural optimization is important, Cohn and
Dinovitzer in 1994, who showed the gap between The Figure 5 above shows the procedure for design
theories and application on designs. The Structural optimization. For a 100 design candidate, the average
Engineering Institute 2002 pointed out the economical performance of population test. After initial
design of RC structures. population, reproduction occurs, wherein the
performance of best candidate may go higher among
The traditional and heuristic are two different the candidate of100. It will reproduce, thus second
approaches, which was separated in two main generation population is obtained. The possible best
branches in the field of structural optimization. candidate in third generation can have same
Genetic algorithm is an effective, heuristic search performance than other generation and so on. Then, it
method (Goldberg, 1989) was used to generate high- will go higher or maintained population. Thus
quality solutions to obtain best performance design generation can reach improved and best performance
and savings generated thereafter. The heuristic as it progresses. The performance of design candidate
approach followed the computerized pathway with the goes up. The best performing design can be obtained
progress of several algorithms, bee colonies, threshold from final generation. These design candidates will be
accepting, stimulated annealing etc. an output.
61
individuals in the current population, called parents,
who contribute their genes—the entries of their
vectors—to their children. The numerical algorithm
was developed in MATLAB software to obtain the
best design performance. The drifts in x – direction is
shown below in Table 2.
Grd 0 0 0 0 0
𝑥𝑡
62
would be possible if the design spectrum satisfies
specific requirements.
This also presents analysis of building structures
using Response Spectrum Method for the design
optimization of building structures. For this study, the
member size of columns and beams were checked for
displacement when the building is under seismic
excitation. The following figure shows the typical
behavior of the genetic algorithm. The whole fitness
assignment, selection, recombination and mutation
process is repeated until a stopping criterion is satisfied.
The evolution process starts with randomly created
first generation from site-specific earthquake records of
Mindanao earthquake records.
(b)
The recorded earthquakes were documented from
the “Operation and Maintenance of Strong Motion Figure 10. Best Fitness Scores per generation
Accelerograph Network of PHIVOLCS last October
2019 and November 2019.
In this study the dynamic analysis using Staad The design candidates are shown in the figure
software was communicated to MATLAB software for below as results of analysis of design using
optimization methods. Once MATLAB software gives STAAD.Pro software. At each step, the genetic
the candidate, the STAAD automatically open to algorithm uses the current population to create the
compute, then results was obtained. Then it will closed. children that make up the next generation. The
STAAD.Pro dynamic analysis is activated once called algorithm selects a group of individuals in the current
by MATLAB software. The program calls directly the population, called parents, who contribute their
results of run analysis of STAAD.Pro of software and genes—the entries of their vectors—to their children.
automatically alters the values of variables. The optimization method was used in this study
through MATLAB optimization software.
The dynamic analysis using response spectrum
III. Results and Discussion method is carried out in this five story building to assess
the seismic behavior when subjected to earthquake
The optimization process of RC beams and loadings. The result of calculated displacements or
columns were carried out in order to search for an each floor level is shown below in Figure 11.
optimum solution, the outputs of the optimization
process, minimization of the displacement versus the
number of generation presented in Figure 10. The
graph shows the plot of best fitness of the average x 5=xt
performance of population test. Result shows the best x5
performance is reached after one hundred generation. 5th Floor
x4
4th Floor
x3
3rd Floor
x2
2nd Floor
x1
1st Floor
63
The inelastic drift according to the NSCP code is REFERENCES
satisfied. The important part of the spectrum that needs
attention is the maximum peak response of strong [1] Atkinson, G.A. and Boore, D.M. Emperical ground
ground motions that might affect the structural motions relations for sub-ductions-zone earthquakes
behavior of buildings. and application to Cascadia and other regions.
Bu,93(4):1703-1729, April 2008.
[2] Boore David M, The Motion of the Ground in
The final displacement/drift response was checked Earthquakes, Scientific American, Vol. 237, No. 6. pp
against code limits. The allowable drift limits are 68-81, December 1977.
satisfied. Thus, building with its structural [3] Chopra A.K. Dynamics of Structures: Theory and
Applications to Earthquake Engineering, Prentice-
configuration such as member sizes, column and beam Hall.1955.
quantities and locations are effectively proportioned. [4] Clough R.W., Penzien J. Dynamics of Structures, 2nd
Edition, McGraw Hill.1994
[5] National Structural Code of the Philippines, 7th edition,
Volume 1, 2015, NSCP C101-10 ISBN 2094-5477,
Association of Structural Engineers of the Philippines
(ASEP), 2015.
[6] FEMA Publication 451, NEHRP Recommended
Provisions for New Buildings and Other Structures:
IV. CONCLUSION Training and Instructional Materials FEMA451B-
June2007
The results obtained from the final generation of [7] Handbook to the Uniform Building Code: an
genetic algorithm progresses until it reach its final Illustrative Commentary. Whittier, Calif.: International
generation. It is found that genetic algorithm is capable Conference of Building Officials, c1988.
for the optimization of structural sizes of reinforced [8] T. Guendelman, G.R. Saragoni, R. Verdugo.
CHILEAN EMERGENCY SEISMIC DESIGN CODE
concrete beams and column wherein the best design is FOR BUILDING AFTER EL MAULE 2010
chosen out of one hundred population, the design EARTHQUAKE 15th World Conference on
performance goes up. Thus, the optimal building design Earthquake Engineering at Lisboa. 2012.
subjected to seismic loading can be achieved to create [9] https://openknowledge.worldbank.org/handle/10986/8
resilient designs and economical designs to protect life 748
and property. [10] M. Ashchheim, E. Hernandez-Montes, L.M. Gil-
Martin. Design of Optimally Reinforced RC Beam,
Column, and Wall Sections. Journal of Structural
Engineering. Vol. 134, No. 2, February 2008.
ACKNOWLEDGMENT [11] I. Rahmanian, Y. Lucet, S. Tesfamariam. Optimal
The authors wish to acknowledge ERDT, DOST, Design of reinforced concrete beams: A review.
Techno-Press, Ltd. Computers and Concrete, Vol. 13,
SEAD and PHIVOLCS for allowing to use the No. 4. 2014
waveform data in this study produced under the project [12] K. Tajziehchi, A. Ghabussi, H. Alizadeh. Control and
entitled “Operation and Maintenance of Strong Motion Optimization against earthquake by using genetic
Accelerograph Network of PHIVOLCS.” MSU- Iligan algorithm. Journal of Applied Engineering Sciences.
Institute of Technology and to those who in one way or Vol. 8(21) Issue 1/2018/
another extend their help for the success of this study. [14] Structural Engineers Association of Davao, Inc.
64
Forensic Wind Hazard Analysis of Typhoon Rolly
Joshua C. Agar1, a * ,Jaime Y. Hernandez Jr., 1,b ,Timothy John S. Acosta, 1,c
1
Institute of Civil Engineering, University of the Philippines, Diliman, Quezon City
a
jcagar@up.edu.ph, bjyhernandez@up.edu.ph, ctsacosta@up.edu.ph,
Abstract—Typhoon Rolly was regarded as the strongest determining the low-pressure field and the wind field of the typhoon.
tropical cyclone in 2020. At peak intensity, it made landfall to the Also, as part of the forensic analysis, the research aimed to perform
Bicol region in the morning of November 1, however the peak forensic structural analysis on documented damages to verify the
estimates from various weather agencies differed, with PAGASA magnitude of the wind hazard.
estimating 10-minute sustained winds of 225 kph and the JTWC
estimating 1-minute sustained winds of 315 kph. Differences in the II. METHODOLOGY
wind speed estimates of various weather agencies have led the
research into performing forensic analysis on both the ground A. Pressure Field Modelling
surface data and the documented damages in order to determine the
magnitude of the wind hazard brought by Typhoon Rolly. As Typhoon Rolly approached landfall, the barometric
pressure drop brought by the approaching typhoon was closely
Keywords—typhoons, forensic analysis, typhoon monitored across the Automated Weather Stations (AWS) of
modelling, severe wind, Monte Carlo simulation, Typhoon Goni, PAGASA. To capture the strength of Typhoon Rolly at the time of
failure simulation, computational fluid dynamics, structural analysis its first landfall at Catanduanes at peak intensity, the AWS on Pili,
Legazpi, Guinobatan, Juban, and Catarman were monitored.
I. INTRODUCTION
Lying on what is called the typhoon belt, the Philippines
is frequented by typhoons by about 20 typhoons a year. Out of 20
typhoons, an average of 5 typhoons make landfall in the Philippines,
bringing hazards like storm surges, floods, and severe winds.
Among the regions frequently hit by strong typhoons was the Bicol
Peninsula. The region's shared history with strong typhoons has the
residents of the region experienced devastating chapters like
Typhoon Sening (International Name: Joan) in 1970 where the
synoptic station in Virac, Catanduanes has recorded then-record
wind speeds of 277.2 kph. It would be soon supplanted by the
current record of 320 kph recorded also on Virac during Typhoon
Reming (International Name: Durian) in 2006.
Fig. 1. PAGASA Automated Weather Stations Monitored
Since Typhoon Yolanda (International Name: Haiyan),
the Philippine Atmospheric, Geophysical, and Astronomical
Services Administration (PAGASA) has introduced Public Storm
Signal #5 to accommodate the newly introduced category called
super typhoons whose 10-minute sustained wind speed was more
than 220 kph.
65
As a result, a scatter plot relating barometric pressure with C. Wind Exposure and Local Winds
respect to the distance of the center of the storm was obtained as
shown in Figure 2.
Fig. 2. Barometric Pressure with respect to distance from the Wind exposure plays a factor in either amplifying or
storm center. reducing the magnitude of the winds as the regional winds are
converted to local winds experienced by the area. Gustiness – a
To normalize the pressure field of Typhoon Rolly in fluctuating increase in magnitude due to turbulence effects,
order to distinguish the pressure drop from the ambient pressure, increases as the wind transited from the sea to land. Shielding occurs
the research adapted Schloemer's hyperbolic model in 1954 [2]: when a feature of land or a huge structure covers the areas on its
leeward side. Topographic amplification occurs when the slope of a
−𝐴
𝑝(𝑟) = 𝑝0 + (𝑝𝑛 − 𝑝0 )𝑒𝑥𝑝( ) (1) terrain causes contraction of the wind flow leading to the increase.
𝑟𝐵
Where p(r) is the barometric pressure at distance r,𝑝0 is Legazpi City, for example, is only exposed to winds
the minimum central pressure of the storm, 𝑝𝑛 (Pa) is the ambient coming from the east, as the Mayon Volcano shields the city from
pressure, and A and B are the shape parameters of the storm. winds coming from west. Thus, Legazpi City is only exposed to
severe winds brought if the typhoon passed over the areas south of
the city. Historically, the city experienced its record maximum wind
B. Regional Wind Field Modelling of 234 kph when Typhoon Sisang (International Name: Nina) passed
south of the city in 1987. Typhoon Tisoy (International Name:
The winds comprising the circulation of the tropical Kammuri) last year, which has a regional wind of 170 kph, traversed
cyclone can be divided into two: winds generated by the primary a similar path as Typhoon Rolly in the Lagonoy Gulf. Virac
circulation and winds generated by the second circulation. The Synoptic Station recorded peak gust of 212.2 kph during the passage
primary circulation is primarily driven by the balance between the of Typhoon Tisoy, while Legazpi Synoptic Station recorded peak
Coriolis forces – which is the fictitious inertial force experienced by gust of 151.2 kph.
a moving body of fluid due to the rotation of the Earth – the
centrifugal force experienced by the rotating body of fluid, and the As Typhoon Rolly traversed across Lagonoy Gulf, the
force caused by the pressure gradient brought by the low pressure eastern and southern coastal towns of Catanduanes and the coastal
center of the storm, whilst the secondary circulation is primarily towns along the coast of Lagonoy Gulf experienced the full brunt of
driven by the heat exchange between the lower and upper the winds.
atmospheres within a cyclone.
Focusing on the primary circulation, the regional winds of D. Forensic Structural Analysis
the typhoon were then determined by achieving the balance between
the pressure gradient forces, the centrifugal force, and the Coriolis Another way to establish the magnitude of the wind hazard
force, which in equation form is known as the gradient wind brought by Typhoon Rolly was through forensic analysis of the
equation: damage incurred by Typhoon Rolly.
𝑈2 1 𝜕𝑝
𝑓|𝑈| + = (2)
𝑟 𝜌 𝜕𝑟
66
Fig. 6. Computational Domain for the CFD Analysis
Fig. 7. Coefficient of Pressures across the building surface Fig. 10. CFD Analysis of CNHS at 45 degrees wrt the structure (top)
and at 30 degrees wrt the length of structure (bottom) shown on the
length.
67
was assessed whether its resistance capacities were exceeded by the
wind load that it was subjected to. For the purlins, assuming that To further eliminate the possibilities about the wind approach, the
they were adequate braced, two scenarios were assessed and wind pressures experienced by the eaves were evaluated also in the
observed: (1) whether the yielding capacity of 248 MPa was CFD analysis. The failure of panels on the eaves, which were
exceeded due to bending or (2) the purlin connection assembly with possible only if negative pressures were experienced, were also
capacity of 232.5 kN was exceeded. For the roof fasteners, which observed at CNHS
were spaced by 0.2 m, the pull-out failure criteria was being assessed
whether each fastener exceeded the tested capacity of 538.7 N with
Coefficient of variance of 0.1. [9]
15 deg 30 deg
45 deg 60deg
68
251 kph, where the first cases of the said failure were documented ACKNOWLEDGEMENT
to have happened. However, this may reflect to possible sub- The main author thanks the PAGASA for providing the
standard construction practice since the structural code suggested historical data for the synoptic stations along with the time history
that the building should withstand wind speeds up to the 320 kph of the data from the AWS during the passage of Typhoon Rolly.
[10].
Collaborating with PAGASA in the Severe Wind Hazard
TABLE I. Failure Simulation Results - CNHS and Risk Assessment program is the UP Institute of Civil
Engineering, under the research project titled "Enhanced Severe
Wind Vulnerability Curves of Key Building Types in the
Angle wrt to Failure Simulation Philippines". Thru the project, the authors are able to collaborate
the length of under this independent research venture.
the structure
[deg] Failure Simulated Lower Upper The main author also acknowledges the Twitter user
Bound Bound @chinitsko for his footage of the failure of school building analyzed
[m/s] [m/s] in this research.
15 Purlin failure 70 81
REFERENCES
30 Purlin failure 63 71 [1] Masters, J., “Super Typhoon Goni Slams Into Philippines As
Strongest Landfalling Tropical Cyclone On Record » Yale
Climate Connections”. [online] Yale Climate Connections.
45 Purlin failure 61 72 2020 Available at:
<https://yaleclimateconnections.org/2020/11/super-typhoon-
goni-slams-into-philippines-as-strongest-landfalling-tropical-
cyclone-on-record/> [Accessed 20 November 2020].
60 Purlin failure 69 74 [2] R. W. Schloemer, “Analysis and synthesis of hurricane wind
patterns over Lake Okechobee”. FL. Hydromet Rep. 31, pp. 49.
75 Purlin failure 71 80 1954
[3] G. Holland, “An Analytic Model of the Wind and Pressure
Profiles in Hurricanes”
69
Environmental Analysis and Estimation of
Evaporation Rate using Pennman Numerical
Simulation in the Salt Production Industry in
Occidental Mindoro
Michelle Enriquez, a *,
michelle_d_enriquez@dlsu.edu.ph
Abstract—Salt or sodium chloride has been in use as a national salt requirement comes from the province with the biggest
flavoring substance to improve the taste; or as preservatives for chunk of the demand supplied by imported salts.
meat, fish and other food products. The salt production is
considered as one of the oldest industry in the world ranging from The province, being an island and surrounded by water,
the traditional solar evaporation to advanced technology-based has one of the longest coastline but is greatly affected by the
harvesting method. Occidental Mindoro was once known in salt climate change and decimated the salt industry. The traditional
production industry contributing almost 18% of the country’s salt method of producing salt used in the province depends primarily
production. But the industry experienced losses, as climate change on solar evaporation is affected adversely by erratic weather
impacts continue to worsen the weather patterns. As the province patterns. The physical and chemical composition of salt produced
is trying to reclaim its reputation in salt industry, salt farmers are varies dependently in many factors: production techniques,
continuously seeking ways on increasing their harvest and ensure processes and the climatic conditions. Salt production through
production year-round and even during the off-season. Historical solar production is mainly dependent on dry season.
analysis is presented in this study to determine which among the
physical, environmental/ecological and biological system has a As the province is trying to reclaim its reputation in salt
great impact on the salt production, characterizing the processes in industry, salt farmers are continuously seeking ways on increasing
the solar production was presented to understand how each stages their harvest and ensure production year-round and even during the
is affected by climatic patterns. Results of the study shows that off-season. Seeing the potential of the salt industry to drive the
meteorological factors like temperature, wind speed, atmospheric local economy of the province, Occidental Mindoro Salt Council
pressure, wind speed and humidity greatly affects the evaporation was created in 2018 which pieced together a roadmap for the
rate. Higher temperature, minimum wind speed, lower humidity Occidental Mindoro Salt industry with the end view of developing
and less rainfall increases the evaporation rate thus speeds up the a salt industry that meets the domestic and international standards
evaporation process and produces more salt). in order to provide equitable benefits by using socially responsible
and environment-friendly practices. The roadmap envisioned that
Keywords— Environmental analysis, evaporation rate, by 2022, the province of Occidental Mindoro shall be known as the
Pennmann- Numerical solution, salt production Salt Capital of the Philippines producing high quality salt products
at competitive prices for the domestic market catering to the
I. INTRODUCTION consumers, food processors, industrial and agricultural users [6].
Salt is considered as one of the important economic As climate pattern in the province changes in time, salt
mineral that is widely distributed on all the continents occurring in industry experiences fluctuation on the volume of salt produced
large reserves [1]. Salt production is one of the most ancient and affecting salt farmers and traders. Longer dry season increases the
widely distributed industries in the world where almost 120 salt production while rainfall and continuous rain flooded the salt
countries have salt producing facilities ranging from primitive solar beds. The objective of this paper is to understand the salt
evaporation to advanced and multi stage [2]. The annual world production processes in the province through characterization of
production of salt can exceed to 200 million tons with more than the physical and environmental factor and determine how climatic
one-third produced by solar evaporation of sea water and inland conditions affect the salt industry.
brines, and the rest by mining of rock salt [3].
Historical analysis is presented in this study to determine
Salt industry in the Philippines was once a vibrant which among the physical, environmental/ecological and
industry. In 1990, the provinces of Bulacan, Pangasinan, biological system has a great impact on the salt production. This
Occidental Mindoro and Cavite supplied almost the 85% of the study aims to present the existing condition of salt industry in
country’s annual salt requirement. In 1994, the country can still Occidental Mindoro using the data obtained in three weather
produce 300,000 metric tons of salt. But in 2010, salt production stations: Barrio Site, Toong and Laste. Specifically, it aims to:
drastically dropped to less than half and significantly resulted to characterize the physical and environment system of the salt farms;
the importing of salt [4]. present the processes involved in salt production; determine the
evaporation rate using the climatic characteristics identified in the
Occidental Mindoro was once one of the biggest salt three sites; and dentify the evaporation rate in the salt production.
producers in the country. In 1990, it produced and contributed
about 60,000 metric tons of the 338,000MT or 18% of the
country’s total salt production [5]. Today, while Occidental
Mindoro still supplies salt in the neighbouring provinces in
Southern Luzon, Visayas, and Mindanao, only 12 percent of the
70
I. METHODOLOGY determine the prevailing characteristics in the salt farms.
Meteorological data from three (3) weather stations: Laste, Sition
A. Study Area Toong and Barrio Site from January 16, 2019 to May 29, 2019 are
used to estimate the evaporation rate on the salt farms.
In the three sites where salt farms and the weather
stations are situated, salt beds are located adjacent to the coast, and A historical background in terms of annual volume of
salt beds are slight higher than the normal high water line. Sitio salt produced is presented as part of the site analysis. The method
Toong and Laste are located in the municipality of Magsaysay, used including the preparation, harvesting and transportation of salt
Occidental Mindoro, both facing the China Sea with vast plains is further identified to describe the processed of salt production.
and valleys. Slope ranges from level to very steep sloping areas;
highest point of elevation is 543 ft. above sea level. The climatic factors like proximity to sea water,
elevation, soil type and other hydrological parameters were
identified to determine the evaporation rate. Numerical Simulation
using the Pennman equation as modified by [9] is used to
determine the evaporation rate for the four-month period.
𝛥 𝛾
𝜆𝐸 = 𝑅𝑛 + 𝑓(𝑢)(𝑒𝑠 − 𝑒) (1)
𝛥+ 𝛾 𝛥+ 𝛾
On the other hand, Sitio Barrio site, is situated in A.2 Stocking of Seawater
Labangan Poblacion, San Jose lying along at nearly 12.3582 lat
and 121.0776 long, and elevation estimated at 9.9m or 32.5feet Stocking of sea water in the reservoir is done through
above mean sea level [8]. dikes connected to rivers or canals. The design of the salt pans
considers the level of tide to ensure that sea water will flow freely
through the gates of the reservoir. The seawater from the reservoir
B. Pennman Evaporation rate flows into the condenser pan. Then water flows through the inlet
canal placed in the condenser to the crystallizer and only high
To characterize the physical and environment system in
saline water is transferred from condenser to the crystallizer, then
thee three sites, an environmental scanning is conducted to
71
the salt crystals start to form. The accumulation of salt increases The table significantly shows that much of the salt
by the rate of evaporation in the environment. It is continuously production is bought by big traders and buyers and only small
evaporated by solar heating and wind flow. As the water portion by TAMACO. As stated by the cooperative, it is because of
evaporates, the salinity rises and salt crystals formed. When the the higher price that the traders offered for each bag of salt. And
water dries up, the salt crystals are now ready to harvest by raking also one of the reason why the cooperative is appealing for
or sweeping. establishing a fix price of salt per bag in the province.
Sea water is stored in the reservoir for almost 2 weeks, The salt harvested in each salt farms brought by Traders,
then evaporated off from the direct sunlight and let the remaining other buyers and TAMACO are stored in salt warehouses and
sea water dense with salt. The gate of the reservoir is then opened transported outside the province. A Salt Washer located in
and allowed to flow in the salt beds. It is then dried up for more Mapaya, Magsaysay was owned and operated by TAMACO that
than 3 hours until salt crystals are seen in the surface. The salt can washed the raw salt to obtain a whiter salt granules free of dirt
crystals are harvested by raking or sweeping. Harvesting of salt is a and other impurities. TAMACO allowed other salt farmers to use
difficult activity since it entails direct exposure to sun. From the the salt washer to ensure that a cleaner salt is transported and
salt farmers in the province, there are almost 4,035 regular salt delivered.
farm workers harvesting the salt produced (TAMACO, 2019).
Figure 2 shows the production rate from 2008 to 2017. A.5 Marketing and the Price Fluctuation
600,000
Price Variation per bag from
400,000
January-May 2019
250.00
200,000
200.00
-
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 150.00
72
B. Factors affecting Salt Production Moreover, the annual meteorological data obtained were
used to determine the average annual evaporation rate. Minimum
B.1 Environmental Condition evaporation rate is during the month when the cloud cover is high
and average wind velocity is near the annual minimum. While,
A thorough environmental investigation is needed to maximum evaporation rate is during the dry season covering the
assess the acceptability of a site for salt production. Assessment is month of March to May when the temperature is high, skies are
barely based on the natural features of the surrounding clear and wind speed is high. Occidental Mindoro experiences high
environment. The salt farm areas in Occidental Mindoro are temperature or dry season during the month of November to April
mainly found in the coastal wetlands. The availability and with high relative humidity and rainy seasons during the month of
accessibility of salt water is first to consider in determining the May to October..
viability for site production. Soil condition is also considers in the
site selection as it affects the soil permeability that determines the Pennman equation of evaporation rate is used in this
capacity of the underlying soil to hold water. Seepage rate is also study to obtain the average annual evaporation rate in the province
consider as it is dependent on the soil type and soil particle size. and it is then compared against the salt production rate. Data from
Soil type in the three sites are sandy clay and silty clay. including the wind speed, humidity and temperature were
simulated using numerical solution. Figure 5 shows the annual
B.2 Seasonality evaporation rate against the production rate of salt industry in the
Climate plays an important role in sea salt production.
The variation in rainfall patterns as affected by the climate change Evaporation rate vs. production
greatly affect the production rate of salt industry. The province’s
erratic rainfall patterns causes flooding in the salt beds. Salt rate
production process was then affected as sea beds were inundated 1,200,000 9
by water. Moreover, temperature also affects the salt production;
the higher the temperature, the more salt can be produced. Figure 4 8
shows the rainfall amount m average temperature and wind speed
1,000,000
from 2009-2018 in Occidental Mindoro. Meteorological data
VOLUME OF SALT PRODUCED (BAGS)
7
during the month of April and May were obtained and considered
in the analysis since evaporation rate is significantly high during
dry seasons. 800,000 6
speed (m/s) 4
3000.0 35
400,000 3
Rainfall amount (mm)
2500.0 30
25 2
2000.0
20 200,000
1500.0 1
15
1000.0
10 Annu
- 0
500.0
al
5 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Produ
0.0 0 ctio…
wind speed Figure 5. Evaporation rate vs. production rate from 2008-2017
73
Table 2. Summary of Evaporation rates in this study, and DOST: PAG-ASA for the meteorological data
used in the simulation.
Location Evaporation rate: Evaporation rate:
min vale (mm/day) max value
(mm/day)
Barrio Site 3.7756 14.2650 REFERENCES
Laste 4.9875 12.2523 [1] Affam M., Asamoah D.N., 2011. Economic Potential of Salt
Toong 9.7789 11.8849 Mining in Ghana Towards the Oil Find. Research Journal of
Environmental and Earth Sciences 3(5):448-456
The table significantly shows the variation of [2] Hossain, Shahadat M., Hossain, Zakir M., Chowdhury
evaporation rates in the three sites. The simulation also proved that Sayedur Rahman. 2006. An Analysis of Economic and
evaporation rates are high during the dry months with maximum Environmental Issues Associated with Sea Salt Production in
values during April and May as minimum wind speed is also Bangladesh and Thailand Cost. International Journal of
observed ecology and Environmental Science 32: 159-172
[3] Sedivy, Vladimir M. 2009. Environmental Balance of Salt
The behavior of evaporation in Laste differs in Toong, it Production Speaks in favor of Solar Saltworks. Global Nest
is because the wind blows directly coming from the coast in Journal 11 (1): 41-48
Toong. Laste, although small in distance with the coastline, is [4] Acedo, Celia E., Hontucan, Royanni Miel. 2017. PH Salt
surrounded by mangroves that block the wind to directly hit the Industry Reeling from Climate Change. Conference of the
salt beds. Further, Barrio Site is situated in the inner part of the Philippines Population Association.
Labangan Poblacion surrounded by almost residential and [5] Business Mirror, 2017. Occidental Midnoro boosting salt
commercial structures. Evaporation rate play an important factor in indsyutry through new technology.
salt production. Temperature, relative humidity, wind velocity, https://businessmirror.com.ph/occidental-mindoro-boosting-
atmospheric pressure can sped up or slows down the evaporation salt-industry-through-new-tech
rate. [6] Department of Trade and Industry. 2018. Salt Industry
Roadmap 2017-2022. www.dti.com.ph
[7] PhilAtlas, 2019. Magsaysay, Occidental Mindoro.
IV. CONCLUSION http://www.philatlas.com
[8] PhilAtlas, 2019. San Jose, Occidental Mindoro.
This study presented the processes in the solar salt http://www.philatlas.com
production existing in the province of Occidental Mindoro. Salt [9] Akridge, Glen D. 2008. Methods for calculating brine
farms are designed considering the ecological characteristics and evaporation rates during salt production. Journal of
the prevailing climatic patterns in the salt farms. The solar salt Archaeological Science 35(2008) 1453-1462
production in the province is consist of reservoir, condenser and [10] Penman, H.L., 1948. Natural evaporation from open water,
crystallizer. Ninety percent (90%) of the total land area is consider bare soil, and grass. Proceedings of the Royal Society, Series
as evaporation ponds and the 10% as the crystallization pond A 193, 120e145.
where the salt is harvested. [11] Tamaraw Salt Producers of Occidental Mindoro, 2019.
Conditions of salt farms in the province of occidental
Salt harvested in the province is either bought by traders, Mindoro.
other buyers and TAMACO and being sold or transported locally
and outside the province. Price of salt per bags varies significantly
on the volume of salt produced, dry and off-season and lack of
policy on fixed price per bags affecting the small scale salt farmers.
ACKNOWLEDGEMENT
74
Simulating the Impacts of Deficit Irrigation on Corn
Yield and Water Productivity Using Coupled DSSAT
CERES-Maize and Aquacrop Models
Marilyn S. Painagan1, a * and Victor B. Ella2,b
1
Department of Agricultural & Biosystems Engineering, College of Engineering and Information Technology, University of
Southern Mindanao, Kabacan, Cotabato.
2
Land and Water Resources Division, Institute of Agricultural and Biosystems Engineering, College of Engineering and Agro-
Industrial Technology, University of the Philippines Los Baños, Laguna.
a*
mspainagan@usm.edu.ph, bvbella@up.edu.ph
Abstract—This study aimed to simulate the effects of deficit irrigation (DI), which is basically defined as the intentional under-
irrigation or intentional under-irrigation on corn yield and water irrigation of crops below full crop water requirements (crop
productivity using DSSAT CERES-Maize and AquaCrop Models. evapotranspiration, ETc) (Fereres and Soriano, 2007). Deficit
irrigation is a well-accepted practice to optimize and increase water
The crop simulation models were calibrated and validated using
use by allowing crops to withstand mild water stress with none or
observed crop production data. Both models performed only marginal decrease in yield and quality (Costa et al., 2007;
satisfactorily during calibration and validation based on standard Geerts and Raes, 2009). Yield reduction is generally expected when
statistical indices such as RMSE, NRMSE, R2, d-stat and NSE. The crops are subjected to deficit irrigation (DI) (Lamm et al., 1995), yet
calibrated and validated models were then coupled and used to a well-designed deficit irrigation (DI) regime can minimize the
simulate the effect of various levels of deficit irrigation on corn impact on yield and still lead to growers’ profitability (Fereres and
yield. Model simulation results showed that a 15 % level of deficit Soriano, 2007; Rodrigues et al., 2013). Deficit irrigation, coupled
with crop simulation models to investigate multiple alternatives,
irrigation results to maximum yield while a 60% level of deficit
have a pivotal role to play in sustainable water development (Wang
irrigation leads to maximum water productivity. Results suggest that et al., 2009). In developing countries such as the Philippines,
maintenance of soil moisture very close to field capacity does not however, no study on deficit irrigation can be found in published
necessarily lead to increased crop yield and water productivity. A peer-reviewed literature much less the application of crop models
net application depth of 77 mm proved to be optimum when a 60% for quantifying the effects of deficit irrigation on crop yield and
deficit irrigation is imposed. Model simulation results also showed water productivity.
that tasseling stage and silking stage until grain filling stage are the Application of crop simulation models is unquestionably
a sound approach in examining the effect of deficit irrigation on crop
most susceptible to water stress. This study demonstrated that deficit
yield and water productivity, particularly for corn production
irrigation is a practical water saving management strategy. The use systems in the Philippines. Corn, particularly sweet corn, is one of
of coupled crop simulation model (DSSAT CERES- Maize and the most important crops that are produced perennially in the
AquaCrop) proved to be a viable tool to support decision making in Philippines. However, sweet corn growers in the Philippines usually
corn production systems in the Philippines, especially when employ surface irrigation applied on a weekly basis if rainfall does
imposing deficit irrigation. not occur, and hence the soil moisture content is maintained near
field capacity. Consequently, farmers tend to use large amount of
irrigation water, which leads to huge water application losses and
Keywords— AquaCrop, Crop simulation, Deficit Irrigation, DSSAT
wastage.
CERES-Maize, Water productivity
In recent years, simulation modelling studies for corn have
been conducted in the Philippines. For instance, Bondad et al. (2015)
I. INTRODUCTION simulated the impact of conservation agriculture on corn yield using
A global challenge for the agricultural sector is to produce DSSAT CERES-Maize model. Lansigan and Salvacion (2015) made
more food with less water (Toumi, 2016) to sustain the rapidly assessments of the impact of climate change on rice and corn yields
growing world population. This problem is due to the increasing using CERES-Maize and Painagan and Guiterrez (2014) conducted
demand of water use in both non-agricultural and agricultural a simulation study of corn yield using the Aquacrop model.
sectors. Nowadays, increasing population, increasing industries, and However, to date no study has been published in peer-reviewed
the adverse effect of changing climatic condition aggravate the literature on the application of both the FAO’s AquaCrop water
problem on water scarcity. productivity model and DSSAT CERES-Maize in simulating the
Water scarcity leads to water stress, which in turn affects impact of deficit irrigation system on corn yield and water
crop growth and productivity in many ways (Ertek and Kara, 2013). productivity in the Philippines.
Inadequate irrigation water application leads to adverse This study aimed to simulate and analyze the effect of
physiological effects like stunted growth, poor quality and reduce deficit irrigation on corn yield and water productivity by coupling
quantity of crop yield. Water scarcity problems also affect water AquaCrop (4.7) and DSSAT CERES-Maize crop simulation
productivity, which is basically defined as the ratio of the mass of models.
crop yield to the amount of irrigation water consumed. Hence,
innovations, strategies, and plans to increase water use efficiency in
crop production systems are necessary.
To address these issues of water scarcity, a number of
water saving technologies and management strategies have evolved
I. METHODOLOGY
through the years. These include an irrigation strategy called deficit
75
To serve as basis for model calibration and validation, field A. Model Calibration
experiments were conducted at the University of the Philippines,
Central Experimental Station, Los Baños, Laguna, Philippines, Calibration of the DSSAT CERES–Maize model was performed
(14.1800 N 121.2500 E) with an elevation of 21 m above sea level, using the observed cultivar-specific coefficients of sweet corn under
from March to May 2018 and from October to December 2018 for Los Baños, Laguna condition using the localized model input data
the first and second cropping seasons, respectively. taken from the October to December 2018 cropping season, when
The climate at the experimental site falls under Type I soil water was optimal for crop growth. It was also the season when
climate based on the Corona classification, with two pronounced plants did not experience stress. For this study, the localized model
seasons, dry from November to April, and wet during the rest of the
input data included the following: daily weather data, parameters of
year. It has an average annual precipitation and temperature of 1,942
mm and 27.1°C respectively. During the year 2018, the driest month the soil profile, local management condition, anthesis day which was
was March with a total rainfall of only 22.5 mm which occurred measured in days after planting (DAP), physiological maturity day
during the first cropping season. Most of the precipitation during this measured in DAP, stem weight at maturity (kg(dm) ha-1), leaf
year fell in the month of December, averaging 450 mm due to the weight at maturity (kg(dm) ha-1) and emergence day measured in
occurrence of strong typhoons. The soil at the experimental site is DAP. In order to fine-tune the crop model, manual iterative process
clay loam classified as Lipa clay loam for the upper 25 cm layer. of adjusting crop genetic coefficient values of sweet corn was done
The site is nearly flat, well drained and has an average water content
until simulated results corresponded to the observed data.
at field capacity and permanent wilting point of 0.32 and 0.14 m3
m-3, respectively. The AquaCrop (4.7) model was also calibrated for local conditions
The climatological input data such as daily maximum and using data taken from the October to December 2018 cropping.
minimum temperature, rainfall, relative humidity, dry and wet bulb Biomass was measured in tons/ha and the percent soil water content
temperature, wind speed, sunshine duration, and vapor pressure for the first 0.25 m depth of the soil profile was used for calibration.
were collected from the National Agrometeorological Station, A trial and error approach was used in adjusting the non-
University of the Philippines Los Baños, located at about 500 meters conservative parameters to minimize the difference between the
from the study area (Figure 1).
simulated and measured data.
For both models, the goodness of fit between simulated and
observed data during model calibration was assessed using standard
statistical indices such as root mean square error (RMSE),
Normalized root mean square error (NRMSE), coefficient of
determination R2 ,Willmott’s index of agreement (d-stat), and Nash-
Sutcliffe model efficiency coefficient (NSE).
76
Water productivity is defined as the ratio of the mass of yield to the Table 2. Statistical measures of the goodness of fit between
volume of water consumed by the crop usually expressed as kg m-3 observed and simulated growth parameters during calibration of
(Geerts et al., 2009): DSSAT CERES- Maize model.
𝑌
Eq.6 𝑊𝑃 = 𝑎
𝐸𝑇𝑎
Where: No. of RMSE NRMSE R2 d-stat NSE
𝑌𝑎 = 𝑚𝑎𝑠𝑠 𝑜𝑓 𝑦𝑖𝑒𝑙𝑑 (𝑘𝑔) dataset (%)
𝐸𝑇𝑎 = 𝑣𝑜𝑙𝑢𝑚𝑒 𝑜𝑓 𝑤𝑎𝑡𝑒𝑟 𝑐𝑜𝑛𝑠𝑢𝑚𝑒𝑑 𝑏𝑦 𝑡ℎ𝑒 𝑐𝑟𝑜𝑝 (𝑚3 )
Silking 3 0.8165 1.60
day day
In this study, water productivity was defined as the grain produced
(DAP)
in kg. per m3 of water used. The numerator was expressed as the
marketable yield, whereas, the denominator was expressed as the
Maturity 9 2.2 24.44
amount of water applied (the sum of rainfall and irrigation).
day days
(DAP)
III. RESULTS AND DISCUSSION Leaf dry 5 115.8 8.49 0.75 0.927 0.736
mass
(kg/ha)
A. DSSAT CERES- Maize model calibration Stem 5 118.61 9.245 0.97 0.992 0.97
During model calibration six genetic coefficient for maize were dry mass
parameterized for local Philippine conditions (Table 1). These six (kg/ha)
maize genetic coefficients were as follows: P1 is the thermal time
from seedling emergence to the end of the juvenile phase; P2 is the Also, the value of genetic coefficient P5 which is the thermal time
extent to which development is delayed for each hour increase in from silking to physiological maturity was estimated based on actual
photoperiod; P5 is the thermal time from silking to physiological observation of maturity during field experiments. Maturity in sweet
maturity; G2 is the maximum possible number of kernels per plant; corn was referred to as the milk stage which occurs during 70 to 80
G3 is the kernel filling rate; and PHINT is the phylochron interval, days after planting, based on sweet corn’s genetic characteristics.
the interval in thermal time between successive leaf tip appearances. During field experiments, maturity stage (milk stage) was observed
beginning 72 days after planting, as maturity progressed sugar will
Table 1. Estimated genetic coefficients of sweet corn under local be converted into starch, therefore, sweet corn was harvested at 75
Philippine conditions. days after planting. P5 values of 463 to 481 degree days with an
average of 472 degree days were the predicted physiological
Variab Unit GSS09 Satur BSS09 Calibrate maturity with a minimum RMSE value of 2.2 days and NRMSE
le 66 sh2 n sh2 77 sh2 d value of 24.44% (Table 2). As a result, a P5 value of 472.0 was used.
Coefficie Mean leaf and stem dry weight obtained from the 15 experimental
nt plots were 1,364 kg ha-1 and 1,283 kg ha-1, respectively. In the
Default sweet corn genetic process of iteration, it was observed that a G2 (maximum possible
coefficient number of kernels per plant) value of 550 and G3 (the kernel filling
P1 Growing 150 180 175 164.0 rate during linear fill stage under optimal conditions) value of 3 gave
degree the minimum value of RMSE which is 115.8 and 118.61 kg ha-1 for
days leaf and stem dry weight, respectively (Table 2). Phylochron interval
(GDD) (PHINT) as well as P2 (extent to which development is delayed for
P2 expressed 0.30 0.30 0.30 0.30 each hour increase in photoperiod) were kept constant, since
as days photoperiodism does not affect the crop’s development.
Calibrated genetic coefficients for sweet corn were compared with
P5 expressed the genetic coefficients of GSS0966 sh2, Saturn sh2 and BSS0977
in degree 680 738 700 472.0 sh2. These sweet corn varieties were pre- calibrated by DSSAT and
days used as default sweet corn genetic coefficients of crop simulation
above a model CERES- Maize. The genetic coefficient values of sweet corn
base calibrated under local conditions are within the range of the default
temperatu values as shown in Table 1.
re of 80C Results showed a high degree of goodness of fit between simulated
G2 number 600 850 500 550.0 and observed data. Dry weight of leaf and stem were predicted with
G3 mg/day 5.50 15 5.0 3.0 R2 of 0.75 and 0.97, d value of 0.927 and 0.992 and NSE of 0.736
PHINT degree 43.0 35 50 43.0 and 0.97, respectively, as shown in Table 2 and Figure 2.
days
77
Figure 2. Simulated and observed dry mass of leaf and stem during Number of plants 185000 25000
calibration of DSSAT CERES- Maize model. per hectare
Crop performance 50 50
under elevated
atmospheric CO2
concentration (%)
Figure 3. Simulated and observed dry above-ground biomass and
Reference Harvest 50 45 soil water content during calibration of AquaCrop model.
Index (HIo) (%)
78
C. Validation of Calibrated Models ear weight at low irrigation application (80% depletion of the
Available Water). This performance of AquaCrop is comparable to
Models were validated by simulating the biomass and fresh ear yield the study of Greaves and Wang, (2016) when they assess AquaCrop
of sweet corn under different levels of deficit irrigation. The in simulating maize growth and productivity under tropical
observed field experiment data from March to May 2018 cropping environment, they reported that the model tended to overestimate
season was used to validate the crop simulation models. Table 5 the biomass accumulated, with the deviation becoming more distinct
shows the simulation result of fresh fruit yield using DSSAT in the DI treatments especially in water stress treatment, and
CERES- Maize models and AquaCrop. Both DSSAT CERES- therefore, they concluded that as water stress intensifies model
Maize and AquaCrop exhibited good performance in simulating the reliability lessens (Greaves and Wang, 2017). Nevertheless, the
fresh fruit yield of sweet corn with R2 values of 0.9529 and 0.8707, models still generally performed satisfactorily based on the standard
respectively (Figure 4 & 5). statistical indices as shown in Table 6 and 7.
C. Model Simulation
Nineteen levels of deficit irrigation, or percent depletion between
field capacity and permanent wilting point, with an interval of 5%
were simulated using the calibrated and validated models. The
depths of irrigation used in the simulations were generated using the
AquaCrop model, which in turn, were used to run the simulation
Figure 4. Observed and simulated values of fresh sweet corn ear using DSSAT CERES-Maize model. In the simulation process
mass using DSSAT CERES-Maize model during model validation deficit irrigation was applied throughout the growing season.
Simulation results for the various levels of DI are shown in Figure
6. Results showed that 15% DI and 20% DI gave the highest grain
yield at an average yield of 5,267 kg/ha and 5,216 kg/ha,
respectively. These results are plausible since crops under 15% DI
and 20% DI did not demonstrate any observable water stress during
the field experiments. Similarly, 10% DI did not show any sign of
water stress, and yet gave relatively lower yield compared to 15%
and 20% DI.
DSSAT CERES- Maize over predicted the fresh ear weight at high
irrigation application (40% depletion of the Available Water). This
result is also similar to the reported performance of CERES- Maize
by Lizaso et al., (2007). The study reported that “the model
consistently over-predicted fresh ear weight across the range of
values examined”. However, AquaCrop also over-predicted fresh
79
Figure 6. Simulated grain yield at different levels of deficit irrigation sign of water stress, and yet gave a relatively lower yield compared
using DSSAT CERES-Maize. to 15% and 20% DI. Results suggest that maintenance of soil
moisture very close to field capacity does not necessarily lead to
Simulation results also showed that 60% DI gave relatively higher increased crop yield. In terms of water productivity, model
yield than 40% DI, 45% DI, 50% DI, 60% DI and 70% DI. On the simulation results showed that 60% DI generate the highest water
basis of crop yield effects alone, model simulation results suggest productivity at 3.24 kg/m3. Among all DI treatments, the 60% DI
that while increasing levels of deficit irrigation or levels of depletion utilized the least amount of water and still obtained a high yield.
generally lead to yield reduction, maintenance of soil moisture very Therefore, a small amount of irrigation application can still lead to
close to field capacity does not necessarily lead to increased crop increased water use efficiency and crop productivity, as long as it is
yield. being applied at the right phenological stage of a crop. Results also
In terms of water productivity, model simulation results showed that showed that there are phenological stages of crops wherein deficit
60% DI obtained the highest irrigation water productivity with 3.24 irrigation has less impact on crop growth and yield. It was found out
kg m-3. In this study, water productivity was measured as the ratio that V3 to V6 stages of corn were not sensitive to water deficit.
of marketable yield and the amount of water used by the crop Furthermore, results showed that growth stages after the grain filling
throughout its growing season. Among all DI treatments, the 60% stage until harvest corn crop are no longer sensitive to water deficit.
DI utilized the least amount of water and still obtained a high yield, Model simulation results also showed that tasseling stage and silking
making it the most productive in terms of water consumption. stage until grain filling stage are stages that are most susceptible to
Moreover, 50% DI and 70% DI obtained the next highest irrigation water stress, which consequently affects the quality and quantity of
water productivity with 2.51 kg m-3 and 2.50 kg m-3, respectively. crop yield.
On the other hand, 40% DI obtained the lowest irrigation water This study demonstrated that deficit irrigation is a practical water
productivity with 1.50 kg m-3. These results showed that a small saving management strategy. Furthermore, the use coupled crop
amount of irrigation water is capable of producing high yield. simulation model (DSSAT CERES- Maize and AquaCrop) proved
Results also showed that there are phenological stages of crops to be a viable tool to support decision making in corn production
wherein deficit irrigation has less impact on crop growth and yield. systems in the Philippines, especially when imposing deficit
It was found that V3 to V6 stages of corn were not sensitive to water irrigation.
deficit. Furthermore, after the grain filling stage until harvest, corn It is recommended to have further data collection and consider
crop is no longer sensitive to water deficit. These results are external parameters to further verify the model.
generally consistent with the findings of other studies (e.g. Denmead
and Shaw, 1960). ACKNOWLEDGEMENT
Under 60% deficit irrigation, different irrigation depths were applied This study was supported by the CHED-PCARI project
at different phenological stage of corn. (Table 8). Results showed entitled “Development of wireless sensor network-based water
that a net irrigation of 77.3 mm that was scheduled during start of information system for efficient irrigation water management in the
tasseling to grain filling stage produced the highest grain yield of Philippines” and the Engineering Research and Development
5009 kg ha-1. Furthermore, a 69.9 mm of net irrigation that was Technology (ERDT) under DOST-SEI. Special acknowledgement
applied during tasseling to grain filling stage produced a grain yield to Dr. Aurelio A. delos Reyes, Dr. Ronaldo B. Saludes and Dr.
of 5006 kg ha-1. In terms of water productivity, an irrigation depth Delfin Suministrado for your support and comments in this study.
of 6.3mm that was applied at the end of tasseling stage to grain
filling stage obtained the highest water productivity of 8.2 kg m-3. REFERENCES
These results highlight the importance of considering the crop’s
phenological stage when scheduling an irrigation. Results also [1] BONDAD, R. M., ELLA, V. B., SALUDES, R. B., REYES,
showed that tasseling stage, silking stage until grain filling stage are M. R., & MERCADO, A. R. 2015. Simulating the impact of
stages that are susceptible to water stress. Stress can lower the crop’s conservation agriculture on corn yield in the Philippines using
yield at high level if it occurs during these stages. the DSSAT CERES-Maize model. The Philippine Agricultural
Scientist, 98(2): 190-201.
Table 8. Irrigation depths, grain yield, water productivity and [2] COSTA, J.M., ORTUÑO, M.F., CHAVES, M.M. 2007. Deficit
irrigation schedule under 60% deficit irrigation. irrigation as a strategy to save water: physiology and potential
application to horticulture. J. Integr. Plant Biol. 49: 1421–1434.
[3] DENMEAD O.T. AND R.H. SHAW. 1960. The effects of Soil
Moisture Stress at Different Stages of Growth on the
Development and Yield of Corn. Agronomy Journal. J-3064:
272-274.
[4] ELLA, V.B. 2019. Annual report for the CHED-PCARI project
entitled "Development of Wireless Sensor Network-Based
Water Information System for Efficient Irrigation Water
Management in the Philippines".
[5] ERTEK A., B. KARA. 2013. Yield and quality of sweet corn
under deficit irrigation. Agricultural Water Management 129:
138– 144. http://dx.doi.org/10.1016/j.agwat.2013.07.012.
[6] FAO Reference Manual, Chapter 2. June 2012 – AquaCrop,
CONCLUSION Version 4.0
[7] FERERES, E., SORIANO, M.A.2007. Deficit irrigation for
Successful imposition of deficit irrigation strategy can be achieved reducing agricultural water use. Special issue on ‘Integrated
if supported with a decision support system like crop simulation approaches to sustain and improve plant production under
models. In terms of yield, model simulation results showed that 15% drought stress’ J. Exp. Bot. 58: 147–159.
DI and 20% DI provide the highest grain yield at an average of 5,267 [8] GEERTS, S., RAES, D., GARCIA, M., MIRANDA, R.,
kg/ha and 5,216 kg/ha, respectively. The 10% DI did not show any CUSICANQUI, J.A., TABOADA, C., MENDOZA, J.,
80
HUANCA, R., MAMANI, A., CONDORI, O. 2009.
Simulating yield response of quinoa to water availability with
AquaCrop. Agron. J. 101: 499–508.
[9] GEERTS, S., DIRK RAES. 2009. Deficit Irrigation as an on-
farm strategy to maximize crop water productivity in dry areas.
Agrl. Water management. 96: 1275-1284.
[10] GREAVES, G.E., & YU-MIN WANG. 2017. Effect of
regulated deficit irrigation scheduling on water use of corn in
southern Taiwan tropical environment. Agricultural Water
Management. 188: 115–125.
[11] GREAVES, G. E., & YU-MIN WANG. (2016). Assessment of
FAO AquaCrop model for simulating maize growth and
productivity under deficit irrigation in a tropical environment.
Water. 8(12): 557. doi:10.3390/w8120557
www.mdpi.com/journal/water
[12] HOOGENBOOM, G., JONES, J.W., PORTER, G.H.,
WILKENS, P.W., BOOTE, K.J., BATCHELOR,W.D.,
HUNT, L.A., TSUJI, G.Y. 2004. DSSAT v.4.0, vol. 1,
Overview. ICASA, Univ. of Hawaii, Honolulu, USA.
[13] HOOGENBOOM, G. 2000. Contribution of agrometeorology
to the simulation of crop production and its applications. Agric.
For. Meteorol. 103: 137–157.
[14] LAMM, F. R., MANGES, H. L., STONE, L. R., KHAN, A. H.,
& ROGERS, D. H. 1995. Water requirement of subsurface
drip-irrigated corn in Northwest Kansas. Transactions of the
ASAE. 38: 441–448.
[15] LANSIGAN, F. P., & SALVACION, A. R. 2015. Assessing
the Effect of Climate Change on Rice and Corn Yields in
Selected Provinces in the Philippines.
[16] LIZASO JI, BOOTE KJ, CHERR CM, SCHOLBERG JMS,
CASANOVA JJ, JUDGE J, JONES JW, HOOGENBOOM G.
2007. Developing a sweet corn simulation model to predict
fresh market yield and quality of ears. J Am Soc Hortic Sci.
132: 415–422.
[17] MAGAIA, E., S. FAMBA, I. WESSTRÖM, R. BRITO, A.
JOEL. (2017). Modelling maize yield response to plant density
and water and nitrogen supply in a semi-arid region. Field
Crops Research. 205: 170–181.
http://dx.doi.org/10.1016/j.fcr.2016.12.025
[18] PAINAGAN, M. S., GUTIERREZ, H. M. 2014. Simulation of
Corn Yield in Carmen, North Cotabato Philippines, using
AquaCrop Model. USM R&D journal. 22(1): 1-8.
[19] RODRIGUES, G.C., PAREDES, P., GONCALVES, J.M.,
ALVES, I., PEREIRA, L.S.2013. Comparing sprinkler and drip
irrigation systems for full and deficit irrigated maize using
multi criteria analysis and simulation modelling: ranking for
water saving vs. farm economic returns. Agric. Water Manag.
126: 85–96, http://dx.doi.org/10.1016/j.agwat.2013.05.005.
[20] Singh, J., Knapp, H. V., & Demissie, M. 2004. Hydrologic
modeling of the Iroquois river watershed using HSPF and
SWAT. Journal of the American Water Resources Association,
Illinois State Water Survey.
[21] TSUJI, G.Y., UEHARA, G., BALAS, S. 1994. (Eds.), Decision
Support System for Agrotechnology Transfer (DSSAT)
Version 3. University of Hawaii, Honolulu, Hawaii.
[22] WANG, Y.-M., NAMAONA, W., TRAORE, S., ZHANG, Z.
2009b. Seasonal temperature-based models for
evapotranspiration estimation under semi-arid conditions of
Malawi. Afr. J. Agric. Res. 4: 878–886.
[23] WILLMOTT, C. J. 1981. On the validation of models. Physical
Geography 2: 184–194.
81
Chi and hypsometric analysis for watershed
prioritization and conservation
Imelida G. Torrefranca1, a *, Roland Emerito S. Otadoy2,3,b, and Alejandro F. Tongco 1,c
1
Engineering Graduate Program, School of Engineering, University of San Carlos – TC, Talamban, Cebu City
2
Department of Physics, University of San Carlos – TC, Talamban, Cebu City
3
Center for Geoinformatics and Environmental Solutions, University of San Carlos – TC, Talamban, Cebu City
a*
imetorrefranca@gmail.com, brsotadoy@usc.edu.ph, caftongco@usc.edu.ph
Abstract— The study presents a rapid assessment tool in development while HI represents the area under this curve and
landform characterization. The method integrates the proxy for expresses the volume of a basin that has not been eroded [13].
steady-state river channel elevation called chi (χ) and the The study explores the integration of the 𝜒 -metric and the
hypsometric analysis using a 30-m SRTM DEM determined using commonly used hypsometric analysis as rapid assessment tools in
TopoToolbox (MATLAB) and further processed in GIS platforms. river basin management and prioritization as applied in the island of
χ gages the stability of drainage divides while hypsometric analysis Bohol, Central Philippines.
quantifies the stages of basin geological development. A quantity
called mean chi difference, 𝜒𝑚𝑑 , is proposed as a measure of the
degree of potential divide mobility. The disaggregation of large
basin into small hydrologic units enables the identification of areas I. METHODOLOGY
of different erosional stages. Highly erosional areas with highly
mobile divides are identified as areas of high priority for The study comprises of two major analysis: the divide stabilty
conservation. analysis with χ and the hyspometric analysis. Both analysis use a
30-m Shuttle Radar Topography Mission (SRTM) digital elevation
Keywords— chi, hypsometry, divide mobility, 𝜒𝑚𝑑 , basin model (https://earthexplorer.usgs.gov).
geological development, TopoToolbox, GIS χ is computed using the DivideTools [14], a series of Matlab-
based tools for analyzing drainage divide stability built on top of
TopoToolbox [15]. Inputs to run the functions are a DEM and a flow
direction grid, and the parameter values for minimum accumulation
I. INTRODUCTION area and theta. As described by [14], the stability of a divide is based
on a criterion wherein the ‘mean of one side of the divide is within
River basins are hydrologically-defined landforms comprising one standard deviation of the mean of the other side’, otherwise a
of drainage divides and stream channel networks. These landforms divide is unstable. The mean chi difference, 𝜒𝑚𝑑 , is proposed to
are previously assumed to be static, however, observations of stream represent the degree of potential divide mobility. It is computed as
capture [1, 2, 3], long-term erosion studies [4] and numerical |𝜒𝑚𝑒𝑎𝑛,1 − 𝜒𝑚𝑒𝑎𝑛,2 | where, 𝜒𝑚𝑒𝑎𝑛 is the mean values, while
modelling [5, 6] demonstrate that basin shape and network topology
subscripts 1 and 2 refer to the sides of the divide section. The 𝜒𝑚𝑑
adjust in response to any perturbations to boundary conditions.
values are classified into low, moderate and high with values of
Geomorphic analysis would give us a glimpse of the dynamic nature
<135, 135 – 255 and >255, respectively.
of these landforms and further provide us guidance on catchment
The hypsometric analysis starts with the delineation and
management process.
parameterization of basins and subbasins using a DEM with a
The 𝜒 – metric is a measure of drainage divide stability
hydrologic model SWAT (www.swat.tamu.edu). Basins/subbasin
fundamentally based on the ‘differential rates of river channel
attributes are used to compute the hypsometric curve and
erosion on opposite sides of a divide’ [7]. 𝜒 is the integral solution
hypsometric integral (HI). The hypsometric curve is a plot of the
to the stream power model, and is expressed as:
𝑚 ⁄𝑛 relative area (a/A) versus the relative altitude (h/H) where a, A, h
𝑥 𝐴
𝜒 = ∫𝑥 ( 𝑜 ) 𝑑𝑥 (1) and H refer to area within the two adjacent contour lines, total area
𝐴(𝑥)
𝑏
of the basin or subbasin, contour interval, and relief of the basin or
where 𝐴 is the upstream drainage area, 𝐴𝑜 is an arbitrary scaling
subbasin, respectively. The value of a/A is 1.0 at the lowest point in
area, 𝑥 is the location of a point along the channel, 𝑥𝑏 is the base
the basin where h/H=0, and it is 0.0 at the highest point of the basin
level, and 𝑚 and 𝑛 are empirical constants. When disequilibrium
where h/H=1. HI is computed as (Zmean- Zmin )/(Zmax- Zmin) where
exists between opposing river channels, the shared drainage divide
Zmean, Zmin, and Zmax refers to mean, minimum and maximum
is set in motion moving from lower to higher 𝜒 area until equilibrium
elevations, respectively.
is reached. 𝜒 - maps provide a rapid visual assessment of divide
stability over a large area and also characterizes the variability of
river network’s topology and geometry in a river basin [8] as they
transmit tectonic and/or climatic signals throughout the landscape III. RESULTS AND DISCUSSION
[9].
Hypsometry quantifies the erosion proneness and stage of A. 𝜒𝑚𝑑 and divide migration
geologic development of a basin [10]. Hypsometric analysis studies The χmd of a divide is a quantitative anomaly between
the distribution of ground surface area, or horizontal cross-sectional opposing channel heads that share that divide. It is proposed to show
area, of a landmass with respect to elevation [11]. The hypsometric the degree of χ differences which can be used as an indicator of the
curve and hypsometric integral (HI) of drainage basins are common potential extent of divide mobility: the higher the χmd, the higher the
tectonic geomorphologic indices [12]. The geometry of the
hypsometric curve classifies a basin into different stages of geologic
82
Figure 1. Divide stability analysis applied in Bohol Island, Philippines. A)
Drainage divide sections colored according to 𝜒𝑚𝑑 values. B) Direction of
potential migration direction and their 𝜒𝑚𝑑 classification.
ACKNOWLEDGEMENT
83
REFERENCES
[1] D. H. Shugar et al., “River piracy and drainage basin
reorganization led by climate-driven glacier retreat,” Nat.
Geosci., vol. 10, no. 5, pp. 370–375, 2017.
[2] L. F. Bloxom and T. J. Burbey, “Determination of the
location of the groundwater divide and nature of
groundwater flow paths within a region of active stream
capture; the New River watershed, Virginia, USA,”
Environ. Earth Sci., vol. 74, no. 3, pp. 2687–2699, 2015.
[3] R. Linkevičiene, “Impact of river capture on hydrography
and water resources: Case study of Ula and Katra
catchments, south Lithuania,” Holocene, vol. 19, no. 8, pp.
1233–1240, 2009.
[4] H. W. Beeson, S. W. McCoy, and A. Keen-Zebert,
“Geometric disequilibrium of river basins produces long-
lived transient landscapes,” Earth Planet. Sci. Lett., vol.
475, pp. 34–43, 2017.
[5] L. Guerit, L. Goren, S. Dominguez, J. Malavieille, and S.
Castelltort, “Landscape ‘stress’ and reorganization from χ-
maps: Insights from experimental drainage networks in
oblique collision setting,” Earth Surf. Process. Landforms,
vol. 43, no. 15, pp. 3152–3163, 2018.
[6] L. Goren, S. D. Willett, F. Herman, and J. Braun, “Coupled
numerical-analytical approach to landscape evolution
modeling,” Earth Surf. Process. Landforms, vol. 39, no. 4,
pp. 522–545, 2014.
[7] J. T. Perron and L. Royden, “An integral approach to
bedrock river profile analysis,” Earth Surf. Process.
Landforms, vol. 38, no. 6, pp. 570–576, 2013.
[8] S. D. Willett, S. W. McCoy, T. J. Perron, L. Goren, and C.
Y. Chen, “Dynamic reorganization of River Basins,”
Science (80-. )., vol. 343, no. 6175, 2014.
[9] K. X. Whipple and G. E. Tucker, “Dynamics of stream-
power river incision model: Implications for height limits
of mountain ranges,landscape response timescales, and
research needs,” J. Geophys. Res., vol. 104, no. B8, pp.
17,661-17,674, 1999.
[10] O. Singh, “Hypsometry and erosion proneness: a case study
in the lesser Himalayan Watersheds,” J. Soil Water
Conserv., no. 2, pp. 53–59, 2009.
[11] A. Strahler, “Hypsometric (Area-Altitude) Analysis of
Erosional Topography,” Bull. Geol. Soc. Am., vol. 63, no.
2, pp. 1117–1142, 1952.
[12] Y. C. Chen, K. Y. Cheng, W. S. Huang, Q. C. Sung, and H.
Tsai, “The relationship between basin hypsometric integral
scale dependence and rock uplift rate in a range front area:
A case study from the coastal range, Taiwan,” J. Geol., vol.
127, no. 2, pp. 223–239, 2019.
[13] R. El Hamdouni, C. Irigaray, T. Fernández, J. Chacón, and
E. A. Keller, “Assessment of relative active tectonics,
southwest border of the Sierra Nevada (southern Spain),”
Geomorphology, vol. 96, no. 1–2, pp. 150–173, 2008.
[14] A. M. Forte and K. X. Whipple, “Criteria and tools for
determining drainage divide stability,” Earth Planet. Sci.
Lett., vol. 493, pp. 102–117, 2018.
[15] W. Schwanghart and D. Scherler, “Short Communication:
TopoToolbox 2 - MATLAB-based software for
topographic analysis and modeling in Earth surface
sciences,” Earth Surf. Dyn., vol. 2, no. 1, pp. 1–7, 2014.
[16] F. Pavano, G. Romagnoli, G. Tortorici, and S. Catalano,
“Morphometric evidences of recent tectonic deformation
along the southeastern margin of the Hyblean Plateau (SE-
Sicily, Italy),” Geomorphology, vol. 342, pp. 1–19, 2019.
84
Automated Furrow Irrigation System for Maize (Zea
mays L. var. saccharata) Production
Sheila G. Griengo1, a *, Armando N. espino Jr. 2, b, Marvin M. Cinense3, c, Romeo B. Gavino 4, d, Vitaliana U.
Malamug5, d, Theody B. Sayco6, d
1
Mindanao State University -Main Campus, Marawi City, Lanao del Sur
23456
Central Luzon State University, Science City of Munoz, Nueva Ecija
a
sittieayeesha@yahoo.com
Abstract - This study aimed to develop an automated to make smart decisions based on accurate and real‐time data (flow
furrow irrigation system to improve irrigation water management in rate, advance rate, automated gates, and other variables). Thus, this
maize (Zea mays L. var. saccharata) production. An Automated study aimed to develop an automated furrow irrigation system to
Furrow Irrigation System (AFIS) was developed and installed in the minimize the water input and human intervention and improve
field during maize production with the following components: application efficiency while satisfying the plant’s needs.
reservoir with motorized offtakes, AFIS microcontroller, soil
moisture sensor, and field advance sensor which was solar-powered
and with wireless data transmission. The field advance sensor was I. METHODOLOGY
located at 60 m from the furrow head for an optimized irrigation
process. A total of 2508 out of 2880 or 87.08 % of the expected data
points from the start of operation in automated blocks were A. Design of the Major Components of AFIS.
transmitted from the soil moisture sensors to the AFIS
microcontroller. Moreover, the soil moisture sensors and field The AFIS has several components such as concrete
advance sensors successfully transmitted all requests for opening reservoir with motorized offtake and distribution lines as a source
and closing the gates, respectively, during the irrigation process. of irrigation water; soil moisture sensor (SMS) to monitor the
Grain yield (14.1 tons/ha), and water use efficiency (43.25 kg/ha- real-time soil moisture status of the field which initiate the
mm) from the produced maize using the automated system were opening of the offtake when the reading stretches below 50%
significantly different compared to those planted using the moisture allowable deficit (MAD); field advance sensor (FAS)
conventional system. Cost analysis of the system resulted in a BCR which monitors the advancing water from the head furrow until it
of 2.37 which infers that the system was viable for maize production. reaches the sensor located before the end of the furrow and
The developed AFIS allowed irrigation processes to proceed with triggers the closing of the offtake ending the irrigation process;
minimum human intervention except for periodic inspections and and, microcontroller which run the embedded program and
routine maintenance. The AFIS also improved water management control the actions and features of the system.
through efficient irrigation scheduling which improved the yield of
maize by 30% compared to the conventional system. B. Test Performance of AFIS
Keywords—furrow irrigation, soil moisture sensor, field The sensors were tested through its capability to transmit
advance sensor, microcontroller, water management the necessary data to the AFIS microcontroller. Lag time and speed
of transmission were also observed. Transmitted data of the soil
moisture sensors were also subjected to reliability tests. It is the
I. INTRODUCTION probability that the software will not fail for a specified time under
specified conditions.
In the Philippines, the most commonly used irrigation
system is surface irrigation which is considered the oldest and most C. Evaluation of Field Performance of AFIS in Maize Production
common method of applying water to croplands. One of the cheap
and probably low-technique methods of surface irrigation is the F1 hybrid sweetcorn (Sweet Victory) was planted in all
furrow irrigation. This irrigation scheme was characterized by low the assigned blocks with 75-day maturity. An area of 8 m x 65 m
application efficiency (45–60%) and caused significant water losses, plot was utilized for both the automated system and the
mainly due to excess deep percolation from the irrigated fields [1]. conventional system. Each block contains 10 furrows each spaced
Rising labor costs, the competition for water and the need to 70 cm apart. All agronomic and plant protection practices were
conserve the environment has intensified the move to more applied uniformly to both treatments (automated and conventional)
advanced systems. An irrigation system should be designed except for the timing of irrigation. Automated block was irrigated
efficiently to uniformly apply irrigation water and maintain adequate based on the actual soil moisture sensor readings while the
soil water for plant growth and production without causing conventional block was irrigated based on farmer’s practice.
excessive water loss, erosion, or water quality impairment [2]. The
modern challenge for improving plant growth and reducing costs
justifies the development of an automated irrigation system that will
minimize the waste of water and reduce labor and monitoring
overhead [3]. Automation is one means of ameliorating the problems
of high labor requirements and low water use efficiency of surface
systems. Automation of the irrigation system will enable the farmer II. RESULTS AND DISCUSSION
85
A. Concrete Reservoir with Automated Offtake C. Field Advance Sensor Assembly
The installed reservoir was constructed using three 24 in. Fig. 3 shows the components of the field advance sensor
diameter x 39.37 in. long, made of concrete circular culvert. Three (FAS) which includes the following: Arduino Nano 3.0, Nano 3.0
(3) culverts were stacked together at 2.5-m high to store water from expansion board adapter, a water level sensor, a relay module, a
the pump for irrigation. The reservoir was equipped with a 3-in voltage regulator and a transceiver. These were all connected to a
automated offtake for controlling the water flow. The offtake was solar power system composed of solar panels with charge controllers
made of a 3/16-in. thick flanged steel plate body with 4 holes (Fig. and battery.
1). A disc valve with steel base rubber seal ring and a 12-V linear Almost all the components for FAS resembled that of the
actuator were connected at the top of the offtake to mechanically soil moisture sensor assembly except for the sensor used. The FAS
operate the device. The offtake was further connected to the assembly utilized a DC3-5V liquid water level module sensor as a
microcontroller to remotely close and open the offtake during the field advance sensor. The advancing water in the furrow was sensed
irrigation process. before it reached the end of the furrow. This also triggered the
termination of irrigation when the water reached the predetermined
location of the sensor. This location was identified through the
furrow performance evaluation before the start of planting. The FAS
was placed 60 meters from the furrow inlet with an approximate cut-
off time of 75 minutes.
86
Table 1. Mean comparison of different variables using t-test
at 5% level of significance.
VARIABLES TREATMENT
Conventional Automated
87
and the rest of the committee members for their invaluable ideas,
time, and patience accorded during the conduct and in doing this
paper.
REFERENCES
[1] Mitchell, A.R., Shock, C.C. & Perry, G.M., (1995). Alternating-
furrow irrigation to minimize nitrate leaching to groundwater.
Conference Proc ‘Clean Water Clean Environment – 21st
Century’, Kansas City, Missouri, ASAE, 1995
[2] Cinense, M. (2010). Modification, Fabrication and Evaluation
of Low Energy Precision Application System for Furrow
Irrigation. Unpublished Master’s thesis. Central Luzon State
University, Munoz, Nueva Ecija, Philippines
[3] Leroux, M. F. And Raghavan, G. V. (2005). Design of an
Automated Irrigation System
[4] Barutcular C, Yildirim M, Koç M, Akinci C, Toptai I, Albayrak
O, Tanrikulu A, & El Sabagh A (2016). Evaluation of SPAD
chlorophyll in spring wheat genotypes under different
environments. Fresenius Environmental Bulletin 25:1258-1266
[5] Alemi, M. H. (1981). The effect of water deficit on yield and
agronomic characters of corn. Field Crop Research. 37(7):
5955
88
Physical and Mechanical Properties Assessment of
Lignocellulosic Rice Straw-based Rigid Polyurethane
Foam for Structural Sandwich Panel Cores
Mary Ann N. Ahalajal1, a *, Arnold A. Lubguban2,b, Roberto M. Malaluan3,c
1
DOE Student, Department of Civil Engineering and Technology, COET, MSU-IIT
2
Professor, Department of Chemical Engineering and Technology, COET, MSU-IIT
3
Professor, Department of Chemical Engineering and Technology, COET, MSU-IIT
a
ann.ahalajal@gmail.com, baalubg@gmail.com, crmmalaluan@yahoo.com
Abstract—This study investigates the physical properties rigid polyurethane foam first, such as its physical and mechanical
such as density and thermal conductivity and the mechanical properties to perform a complete structural analysis and design. This
properties such as compressive strength and shear strength of investigation was made to determine the properties of the
lignocellulosic rice straw-based rigid polyurethane foam. These lignocellulosic rice straw-based rigid polyurethane foam such as its
properties were further assessed using ASTM E 1730-02 Standard density, thermal conductivity, compressive strength, and shear
Specification for Structural Sandwich Panel Core. Further, the strength based on American Standard Testing Method (ASTM) and
results were compared to petroleum-based rigid polyurethane foam. to assess if the properties meet the requirements for structural
sandwich panel cores as specified in ASTM E 1730-02. The results
Keywords—rice straw-based rigid polyurethane foam, were compared to the properties of petroleum-based rigid
bio-based rigid polyurethane foam, structural sandwich panel core polyurethane foams.
89
INV® 690 (surfactant) 1.5 1.5 accordance with the ASTM C 518-15, by using samples of 200 x
Distilled water 1.0 1.0 200 x 35 mm.
(blowing agent)
A-side materials C.2 Mechanical Properties
Isocyanate indexb of PAPI® 27 110 110
a
C.2.1 Compressive Strength
The concentrations of all ingredients are expressed in parts per hundred part of polyol,
which conventionally dictates that the sum of all polyols adds up to 100 parts.
b
The amount of isocyanate index used relative to the theoretical equivalent amount. The compressive strength of the foam was determined in
The isocyanate index is the ratio of the actual amount of isocyanate used over the accordance with ASTM D1621 by a Shimadzu Universal Testing
theoretical amount of isocyanate required times 100.
Machine with Trapezium X material testing software. The test
specimens were square of cross-section 50 mm x 50 mm and a
B. Rigid Polyurethane Foam Production minimum height of 25.4 mm and loaded at a constant rate of 2.5
mm/min for every 25.4 mm height. The loaded ends of the specimen
Standard laboratory mixing and pouring procedures for are parallel to each other and perpendicular to the sides.
making water-blown polyurethane foams were performed for this
study as used by Tu, et. al (2007). The petroleum-based polyol
(VORANOL 490), rice straw-based polyol, catalysts, surfactant, and C.2.2 Shear Strength
blowing agent (B-side materials) were added into a 3-liter plastic
container and mixed at 2000 rpm for 10-15 seconds. The mixture The shear strength of the foam was determined in
was then allowed to degas for 120 seconds. Methylene diphenyl accordance with ASTM C 273 and ISO 1922 by a Shimadzu
diisocyanate was then added rapidly into the mixture. Stirring Universal Testing Machine with Trapezium X material testing
resumed at the same speed for another 15 seconds. The mixture was software. The test specimen had a thickness equal to 25 mm, a width
then poured immediately into a wooden mold with dimensions 18 of 50 mm, and a length of 250 mm and loaded at such a constant rate
cm x 18 cm x 30 cm covered in aluminum foil lining and allowed to that the maximum load will occur within 3-6 minutes. A constant
rise freely at ambient conditions. The foam was then demolded and rate of 0.7 mm/min was applied.
allowed to cure for 7 days. After curing, the foam block was cut into
appropriate sizes. Fig. 1 illustrates the foaming procedures. D. ASTM E 1730-02 Standard Specification for Rigid Foam for
Use in Structural Sandwich Panel Cores
Petroleum-based Table 2 shows the standard specification for rigid foam for
polyol use in structural sandwich panel cores as listed in ASTM E 1730-02.
This specification is used as a reference in assessing the physical and
Blowing agent (H2O)
mechanical properties of the lignocellulosic rice straw-based rigid
Catalyst
polyurethane foam.
Rice straw-based Surfactant
polyol
TABLE II. PHYSICAL AND MECHANICAL PROPERTIES OF
RIGID FOAM FOR USE IN STRUCTURAL SANDWICH
PANEL CORE
Requirement Type 1 Type 2 Type 3
Procedure
Density, kg/m3, 41.6 55.7 72.0
Polymeric max
Transfer to mold
diisocyanate Thermal 0.036 0.037 0.037
conductivity,
W/mK, max
Fig. 1 Procedure for preparing rigid PU foams
Compressive 172.4 275.8 448.2
strength, kPa,
C. Foam Characterization min
Shear strength, 172.4 241.2 275.8
C.1 Physical Properties kPa, min
C.1.1 Density
The density of the foam was measured in accordance with III. RESULTS AND DISCUSSION
ASTM D 1622-08. Density was determined by dividing the weight
The following presents and discusses the results obtained
of each sample by its corresponding volume. The samples were in performing the methodology in this study.
square in cross-section with an area of 25 cm2 and a height of 30
mm.
A. Physical Properties
A.1 Density
C.1.2 Thermal Conductivity
Table 2 shows the density test results of PU foams in this
The thermal conductivity measurements of the foams study. It can be observed that the average density of control PU
were performed using a heat flow meter. The measurements were foams is higher than the average density of the LRS PU foam
made under steady heat flow conditions through the test samples, in samples. Also, the density results of LRS PU foam have no
significant difference from each other.
90
TABLE II DENSITY OF PU FOAM SAMPLES
Density of PU Foam Samples
Sample Density, kg/m3
LRS PU foam Control PU foam 90
1 55.53 48.88
80
2 55.47 48.76
Density (kg/m3)
3 53.56 50.00 70
4 55.98 80.39 60
5 56.05 77.85
50
6 54.86 75.02
7 56.76 85.36 40
8 56.92 85.23
30
9 56.73 84.93
Standard Type 1 PU Foam, max Standard Type 2 PU Foam, max
Average 55.54 70.71 Standard Type 3 PU Foam, max LRS PU foam
Conrol Pu foam
0.020
TABLE III THERMAL CONDUCTIVITY OF PU FOAM
SAMPLES
0.010
Thermal Conductivity, W/mK Standard Type 1 PU Foam, max Standard Type 2 and 3 PU Foam, max
LRS PU foam Control PU foam LRS: Perp. to Rise Direction LRS: Par. to Rise Direction
Control: Perp. to Rise Direction Control: Par. to Rise Direction
Sample Perp. to Par. rise Perp. to Par. to rise
rise direction rise direction Fig. 3 Thermal conductivity of PU foam samples in comparison
direction direction with the standard thermal conductivity specifications
1 0.03121 0.02786 0.03341 0.04339
2 0.03992 0.03526 0.03986 0.03385 B. Mechanical Properties
3 0.03952 0.03531 0.04095 0.03786 B.1 Compressive Strength
Ave. 0.03688 0.03281 0.03807 0.03837
Table 4 shows the compressive strength results of the PU
foam samples. It is noted that PU foams made with 100% Voranol ®
490 (control PU foam) have higher average compressive strength
Fig. 3 illustrates the thermal conductivity results found in compared to LRS PU foam.
Table 3. These values were compared with the maximum thermal
conductivity values for all types (Type 1, Type 2 and Type 3) of PU TABLE 4 COMPRESSIVE STRENGTH OF PU FOAM
foams used in structural sandwich panel cores as indicated in ASTM SAMPLES
E 1730-02, which is shown in solid line in Fig 2. It can be seen that
the LRS PU foam passed the maximum standard value of 0.037 Sample Compressive Strength, kPa
W/mK. LRS PU foam Control PU foam
1 495.471 414.491
2 544.642 433.798
3 529.328 436.515
4 422.609 638.281
5 461.992 824.466
6 429.083 749.656
7 468.014 893.057
8 476.633 829.393
9 497.645 875.609
Average 480.601 677.252
91
Fig. 4 illustrates the compressive strength test results
Shear Strength of PU Foams
found in Table 4. It can be seen that the LRS PU foam passed all
minimum standard compressive strength values specified in ASTM 5.0
E 1730-02.
3.0
Compressive Strength of PU Foams
2.0
1000
900 1.0
Compressive Strength (kPa)
800 0.0
700
Standard Type 1 PU Foam, min Standard Type 2 PU Foam, min
600 Standard Type 3 PU Foam, min LRS PU foam
500 Conrol Pu foam
400
300 Fig. 5 Shear strength of PU foam samples in comparison with the
200 standard shear strength specifications
100
0
IV. CONCLUSION
Standard Type 1 PU Foam, min Standard Type 2 PU Foam, min
Standard Type 3 PU Foam, min LRS PU foam The results of the study clearly show that the physical
Conrol Pu foam properties and mechanical properties, such as density, thermal
conductivity, compressive strength and shear strength, of the
Fig. 4 Compressive strength of PU foam samples in comparison lignocellulosic rice straw-based rigid polyurethane foam passed the
with the standard compressive strength specifications specifications for structural sandwich panel core listed in ASTM E
1730-02. Further, it can be classified as a Type 3 structural sandwich
panel core.
B.2 Shear Strength This study shows that the lignocellulosic rice straw-based
polyurethane foam can be used in actual applications such as
Table 5 shows the shear strength test results of the PU
sandwich wall panels, roof panels, and others.
foam samples. Shear strength test of the control PU foam yet to be
completed. The average value is calculated from the initial test
results. It can be seen that the average shear strength value is higher ACKNOWLEDGEMENT
than the LRS PU foam.
This study was made possible by BioProducts Research
TABLE 5 SHEAR STRENGTH OF PU FOAM Laboratory, MSU-IIT, Philippines.
SAMPLES
REFERENCES
Sample Shear Strength, MPa3
LRS PU foam Control PU foam [1] “Introduction to Polyurethanes: Polyurethane Applications”,
1 - 2.298 https://polyurethane.americanchemistry.com/Applications/.
2 2.640 2.188 [2] Enderus N.F. and Tahir, S.M. (2017), “Green Waste Cooking
3 2.116 - Oil-based Rigid Polyurethane Foam,” IOP Conf. Series:
4 2.152 3.37 Materials Science and Engineering.
5 2.327 3.292 [3] Icalina, K.M. and Lubguban, A.A. (2018), “Development and
6 2.352 4.453 Optimization of the Physical Properties of Water-Blown Rigid
7 3.175 - Polyurethane Foams Produced from Rice Straw-Based
Polyols,” Material Science and Engineering, MSU-IIT, Iligan
8 3.061 -
City, Philippines.
9 3.097 -
Average 2.615 3.120
92
Silanated Coconut Oil-Based 2K Waterborne
Polyurethane Coating for Corrosion Protection
Dave Joseph E. Estrada1,2 a *, Amierson C. Tilendo2,b, Roberto M. Malaluan2,3,c Arnold C. Alguno4,d,
and Arnold A. Lubguban2,3,e
1
College of Engineering Graduate School, MSU – Iligan Institute of Technology
2
Bio Products Research Laboratory, MSU – Iligan Institute of Technology
3
Department of Chemical Engineering and Technology, MSU – Iligan Institute of Technology
4
Department of Physics, MSU – Iligan Institute of Technology
a
davejoseph.estrada@g.msuiit.edu.ph, bamierson.tilendo@g.msuiit.edu.ph, croberto.malaluan@g.msuiit.edu.ph
d
arnold.alguno@g.msuiit.edu.ph, arnold.lubguban@g.msuiit.edu.ph
Abstract— This study aims to develop novel techniques that coconut oil can also be utilized in synthesizing polymeric
on the synthesis of an effective two-component (2K) coconut oil- materials despite the challenges in its molecular structure.
based waterborne polyurethane with varying silane loadings applied But, as stated earlier that due to the high hydrophobicity
on a carbon steel substrate as protective coating against metal of coconut oil, water-borne approach is not applicable because it will
corrosion. Electrochemical analysis shows a significant decrease on lead to the phase separation between oil and water. Several
corrosion rate as silane loading ratio increases. Moreover, Tape Test researches were conducted to address this concern, and the use of
method is utilized to determine the adhesion strength of the silane is quite conventional. Silane is a silicon containing
synthesized coating material. Results revealed that the adhesion compound. One silicon molecule can have both organic and
strength of the synthesized WBPU material is enhanced as silane inorganic reactivity, thus silane is widely used as adhesion
loading increases. promoters and coupling agents. According to Lyu et al. (2019),
inorganic Si-O bond in silane results in low surface energy, good
Keywords—Waterborne, coconut oil, silane, adhesion thermal stability and excellent flexibility [11].
strength, electrochemical analysis.
In this study, the coconut oil will be utilized as a partial polyol
I. INTRODUCTION replacement for petroleum based polyol for the synthesis of
silanated coconut oil-based 2K waterborne polyurethane coating for
Metallic substrates are susceptible to corrosion attacks in corrosion protection. Furthermore, the addition of silane at different
natural environments and there are so many ways on preventing it loadings will be investigated if it contributes to its corrosion rate
but generally, the use of coating is the most common one [1], [2]. adhesion strength properties.
Organic coating such as the use of environmentally friendly
polyurethane (PU) (e.g., waterborne polyurethane, WBPU) coatings
for corrosion protection have garnered attention over the past couple II. METHODOLOGY
of years, as certain pioneering researchers have improved the PU
coating. PU is mainly composed of polyol (called the soft segment) A. Materials
and diisocyanate (called the hard segment) [3]. The coconut polyester polyol (CCP) was supplied by the
The major components for the synthesis of PUs are Bio Products Research Laboratory, MSU – IIT. Additives such as
produced from petroleum oil as the feedstock [4]. With the polyether-modified siloxane (BYK 349), polyether block copolymer
fluctuation of prices for petroleum feedstock, this dangerously (DISPERBYK-190), foam-destroying polysiloxane (BYK 022),
affects the price of raw materials for industries [5]. Moreover, non-ionic urethane rheology modifier (ACRYSOL RM-8W), and
Desroches et al. (2012) emphasized that the instability regarding the titanium dioxide (Ti-Pure R960) were supplied by supplied by
depletion of fossil based assets and worldwide temperature Chemrez Technologies. The tetraethoxysilane used in this
alteration issues were identified [6], [7]. Renewable resource such experiment was reagent grade.
as natural oil-based polyols are widely used to prepare PUs. For
example, soybean oil, palm oil, castor oil, and rapeseed oil have
higher unsaturated fatty acid content, hence they can be hydroxyl B. Methods
functionalized easily using various methods. However, some of
these nature oils can be produced in small scale and will compete The NCO:OH ratio was kept constant at 1.2:1. The
between food industrials and petrochemical industrials [4], [8]. additives and pigments, including the isocyanate and polyol will be
Coconut oils are abundantly available plant oils and can measured in terms of grams that would account to a 15 g WBPU
produce an appreciable amount of coconut oil up to 60-65 % of its sample. The first step is to disperse the distilled water, defoamer,
actual weight [9]. But in comparison to other nature oils, it has a low dispersing agent, rheolate, and pigment using and overhead stirrer
unsaturated fatty acid content making it difficult to use in PU for 20 minutes at 3000 rpm. Add distilled water and continuously
synthesis. Fortunately, this problem was addressed by Ampon et al. stir for 15 minutes at 1500 rpm. Add the bayhydrol, CCP, distilled
(2019), they have successfully developed a highly functionalized water, defoamer, wetting agent, and rheolate and stir for 20 minutes
coconut oil-based polyester polyol (CCP) via polycondensation at 2000 rpm. After 20 minutes, add an additional distilled water and
process [10]. The appreciable change in the properties of the coconut continuously stir for 3 minutes at 2000 rpm. Lastly, add the HDI and
oil after the polycondensation process provided baseline information continuously stir at 2000 rpm until a thick consistency is achieved.
93
Add silane to the synthesized WBPU with 0-5wt% concentration potential (Ecorr), corrosion current density (icorr), cathodic slope
and stir for 2-3 minutes at 2000 rpm. The parts of all the reagents for (βc), anodic slope (βa) and corrosion rate (CR) which are calculated
each steps are indicated in the Table 1 below: and tabulated in Table 2. The corrosion rate is calculated by using
(1):
0.13 𝐼𝑐𝑜𝑟𝑟 (𝐸. 𝑊. )
𝐶𝑅 (𝑚𝑚𝑝𝑦) =
𝑑
TABLE 1. FORMULATION AND PARTS FOR THE It can be observed in Fig. 1 that tafel plots shifted to lower
SYNTHESIS OF PU COATING current density as the silane loading increases. It indicates that the
Steps Function/supplier Parts incorporation of silane on the improvement of corrosion resistance
Local 5-6 with Si-O-Si act as an inhibitory effect additive with WBPU coating.
Defoamer/BYK 1-2 In potentiodynamic polarization curve analysis, the increase in Rp
Dispersing and decrease in Icorr represented the better protection of corrosion.
1 1-2
agent/BYK
Rheolate/DOW 0.1-1.0
Pigment/Dupont 20-25
2 Local 1.0-2.0
Resin/Bayer 26.0-27.0
Polyol/BPRL 26.0-27.0
Local 0.9-1.0
3
Defoamer/BYK 0.1-0.5
Wetting agent/BYK 0.1-0.5
Rheolate/DOW 0.1-0.5
4 Local 3.0-5.0 Fig. 1. Electrochemical polarization curve test results.
5 Isocyanate 32.0-35.0
The results in Table 2 showed that increasing of silane
loading up to 1 wt% increased the Ecorr and decreased the Icorr values.
C. Application of the synthesized WBPU Since corrosion rate is directly proportional to Icorr value [12], it can
Place the metal substrate on the spin coater. On top of the be deduced that higher silane loading improved the anticorrosive
metal substrate, place the synthesized PU and let the spin coater to properties of the coatings. To further support this finding, the
coat the metal substrate at 1000 rpm for 30 seconds. Repeat calculated corrosion rate (CR) of the WBPU coatings also showed a
application until the coating is uniformly distributed on the metal declining trend upon increasing the loading of silane. The improved
substrate. Curing time must take for at least 1 week. corrosion rate could be due to the fact that silane-based coatings are
chemically inert and can provide corrosion protection because of
D. Characterizations and specifications their good barrier properties that delay the entry of aggressive
species to the metallic surface due to its hydrophobic property as
The corrosion resistance of waterborne PU coatings is
evaluated using potentiodynamic polarization using stated by Asadi et al. (2014) [13].
Potentiostat/Galvanostat (SP150, Bio-Logic). To examine the
adhesion of coatings with the substrate, cross cut adhesion test of TABLE 2. POLARIZATION PARAMETERSFOR THE
prepared coating panels was carried out using the ASTM D-3359- WBPU COATING WITH DIFFERENT SILANE LOADINGS
02 standards. Thermal degradation properties of the polyurethane
foam samples were studied between -30°C to 400°C using DSC Ecorr Icorr CR
Samples βa βc
(DSC 4000, Perkin Elmer) under nitrogen atmosphere at 20 L/min (mV) µA (mmpy)
with a ramp rate of 10°C /min. Chemical compositions of the 0% -589.79 13.765 193.2 382.4 0.029
synthesized PU coating will be evaluated by Fourier Transform 0.2% -468.02 0.466 231.3 262.7 0.98e-3
Infrared (FTIR) using Shimadzu IRTracer-100 ATR mode at 0.4% -496.84 0.371 150.0 776.2 0.78e-3
wavelengths between 4000 and 400cm−1. 0.6% -745.39 0.063 68.9 201.2 0.15 e-3
1% -575.01 0.071 141.4 82.9 0.13 e-3
E. Statistical analysis
B. Adhesion strength test
All experiments are repeated thrice. Therefore, the
experimental results will be on average values from two replications The photographs of all samples that have undergone tape
with standard deviations on the error bars. The imprecision, test method can be seen in Fig. 2. It can be clearly seen that more
coefficient of variation (CV) will be calculated to be < 5 %. In areas are removed during the tape test at 0% loading of silane.
addition, analysis of variance (one-way ANOVA) will be carried out However, as the silane loading increases, the area removed tends to
by Microsoft Excel tool package within 95% confidence interval. decrease.
94
ACKNOWLEDGEMENT
The researchers would like to thank the following
0.2%% sponsors: Chemrez as the official supplier of all reagents used, and
Bio Products Research Laboratory – MSUIIT for allowing the
student to conduct its experiment.
REFERENCES
[1] P. Y. Tsai, T. E. Chen, and Y. L. Lee, “Development and
characterization of anticorrosion and antifriction properties
0.4%
for high performance polyurethane/graphene composite
coatings,” Coatings, vol. 8, no. 7, 2018, doi:
10.3390/coatings8070250.
[2] O. S. I. Fayomi, I. G. Akande, and S. Odigie, “Economic
Impact of Corrosion in Oil Sectors and Prevention: An
Overview,” J. Phys. Conf. Ser., vol. 1378, no. 2, 2019, doi:
0.6% 10.1088/1742-6596/1378/2/022037.
[3] M. M. Rahman, M. H. Zahir, M. B. Haq, D. A. Al Shehri,
and A. M. Kumar, “Corrosion inhibition properties of
waterborne polyurethane/cerium nitrate coatings on mild
steel,” Coatings, vol. 8, no. 1, 2018, doi:
10.3390/coatings8010034.
1% [4] T. Kattiyaboot and C. Thongpin, “Effect of Natural Oil
Based Polyols on the Properties of Flexible Polyurethane
Foams Blown by Distilled Water,” Energy Procedia, vol.
89, pp. 177–185, 2016, doi: 10.1016/j.egypro.2016.05.024.
Fig. 2. Photographs of all samples after tape test. [5] B. J. Rashmi, D. Rusu, K. Prashantha, M. F. Lacrampe, and
P. Krawczak, “Development of water blown bio-based
thermoplastic polyurethane foams,” Adv. Mater. Res., vol.
Figure 3 shows the average remained area after the tape
584, pp. 361–365, 2012, doi:
test is conducted. An increasing trend for its adhesion strength can 10.4028/www.scientific.net/AMR.584.361.
be observed as the silane loading ratio increases. The average area [6] M. Desroches, M. Escouvois, R. Auvergne, S. Caillol, and
remained with standard deviation for samples 0%, 0.2%, 0.4%, B. Boutevin, “From vegetable oils to polyurethanes:
0.6%, and 1% silane loadings are 48±23.25mm2, 62±10.19 mm2, Synthetic routes to polyols and main industrial products,”
74.66±27.76 mm2, 94±3.74 mm2, and 98.33±0.47 mm2, respectively. Polym. Rev., vol. 52, no. 1, pp. 38–79, 2012, doi:
The 1-way annova is calculated and the p value is 0.047 which is a 10.1080/15583724.2011.640443.
[7] S. Sahoo, S. Mohanty, and S. K. Nayak, “Biobased
bit lower than the threshold value of 0.05. This indicates that there polyurethane adhesive over petroleum based adhesive: Use
is a significant difference in terms of its adhesion strength as the of renewable resource,” J. Macromol. Sci. Part A Pure
silane loading increases. Appl. Chem., vol. 55, no. 1, pp. 36–48, 2018, doi:
10.1080/10601325.2017.1387486.
[8] F. Carré, “1. Introduction,” L’archéologie en Haute-
Normandie, Bilan des connaissances, pp. 7–12, 2018, doi:
10.4000/books.purh.5104.
[9] M. Kumar and R. Kaur, “Effect of Different Formulations
of MDI on Rigid Polyurethane Foams based on Castor Oil,”
vol. 2, no. May, pp. 29–42, 2013.
[10] B. S. Ampon, J. J. C. Budiongan, J. D. Mangubat, and B.
A. N. Semense, “Optimum Functionalization of Coconut
Oil-Based Polyester Polyol (CPP) via Polycondensation
Fig. 3. Average remained area of all samples. with Phthalic Anhydride,” Mindanao State University -
Iligan Institute of Technology, 2019.
[11] J. Lyu et al., “In situ incorporation of diamino silane group
into waterborne polyurethane for enhancing surface
hydrophobicity of coating,” Molecules, vol. 24, no. 9, 2019,
IV. CONCLUSION doi: 10.3390/molecules24091667.
[12] M. M. Ariffin, M. M. Aung, L. C. Abdullah, and M. Z.
It can be concluded that it is indeed possible to Salleh, “Assessment of corrosion protection and
synthesized coconut oil-based waterborne polyurethane as performance of bio-based polyurethane acrylate
protective coating against metallic corrosion. In agreement with incorporated with nano zinc oxide coating,” Polym. Test.,
the corrosion current densities obtained from polarization vol. 87, no. April, p. 106526, 2020, doi:
curves, the 1 wt% tetraethoxysilane loading exhibits the best 10.1016/j.polymertesting.2020.106526.
result in terms of lower cower corrosion rate. Furthermore, [13] N. Asadi, R. Naderi, M. Saremi, S. Y. Arman, M. Fedel,
the adhesion strength of the 1 wt% tetraethoxysilane loading and F. Deflorian, “Study of corrosion protection of mild
steel by eco-friendly silane sol-gel coating,” J. Sol-Gel Sci.
WBPU coating shows a significant increase.
Technol., vol. 70, no. 3, pp. 329–338, 2014, doi:
10.1007/s10971-014-3286-8.
95
Qualitative risk assessment of tailings dam
Cielo D. Frianeza1, a* and Andres Winston C. Oreta1, b
1
De La Salle University Manila, 2401 Taft Ave, Malate, Manila, 1004 Metro Manila
a
cielo_frianeza@dlsu.edu.ph, bandres.oreta@dlsu.edu.ph
Abstract—By-products from the mining process, called There is no publicly accessible inventory of tailings dams
tailings, are composed of uneconomical ore, process water, and [18]. Hence, most of the data from studies involving tailings dam
hazardous substances that are usually stored in structures made of come from recorded tailings dam failure. Tailings dam failure
soil called tailings dams. Despite the importance of tailings dam impacts the environment severely due to the heavy debris and
stability because of the hazardous materials it stores, failure often hazardous substances that are present in the tailings. Groundwater,
occurs, which severely impacts the environment, specifically surface water, vegetation, and soil are at risk of contamination and
groundwater, surface water, vegetation, and soil. Thus, the study destruction if tailings are discharged onto them. Thus, the study aims
aims to assess risk qualitatively using the preliminary qualitative to assess the risk to protect tailings dams from failure. However, data
checklist rating and failure modes and effects analysis (FMEA). about tailings dams is limited and the only known tailings dams in
Thus, the hazard occurrence, vulnerabilities of the tailings dam, and the Philippines are those that have already failed. Research has
the factors that contribute to the vulnerabilities were identified to shown that there is a lack of publicly accessible qualitative or
accurately conduct the qualitative risk assessments. Prior to a quantitative risk assessments for active and inactive tailings dams in
detailed assessment, a preliminary qualitative checklist rating was the Philippines. To show the application of qualitative risk
conducted wherein ratings were assigned to the vulnerabilities, assessment, the study applied the preliminary qualitative checklist
hazards, and occurrence of hazards to determine whether a detailed rating and failure modes and effects analysis (FMEA) on an active
analysis of the tailings dam is required based on the level of risk. tailings dam that have failed in the past. The study will also
Since Padcal Mine Site Tailings Pond 3 is a high-risk facility, a more determine the hazard occurrence in areas wherein operational mines
detailed risk assessment, FMEA, was applied wherein the are situated, assuming that the tailings dam of the respective
vulnerabilities of the said tailings dam were evaluated to determine operational mine is nearby as a reference for future risk assessments.
which component will cause failure. Based on the analysis,
overtopping, inadequate foundation design, inadequate structural II. CONCEPTUAL FRAMEWORK
design, and mine subsidence are the probable cause of failure of
Padcal Mine Site Tailings Pond 3 due to the frequent typhoon and The conceptual framework of the study is shown in Fig. 2.
earthquake incidence, as well as the existing condition of the facility. The study will focus on risk assessment rather than the other
components of disaster preparedness. Furthermore, to assess the
Keywords—Tailings dam failure, Qualitative risk risk, the hazard occurrence, element at risk, and vulnerability of the
assessment, Failure modes and effects analysis, Environment element at risk must be determined. Since the common natural
hazards that trigger tailings dam failures in the Philippines are
I. INTRODUCTION rainfall and earthquakes, the occurrence of these hazards on the
operational metallic and non-metallic mines in the Philippines will
Tailings is a slurry composed of the uneconomical part of be determined using hazard maps. On the other hand, the study will
the ore, excess water used during the process of separating the focus on common vulnerabilities of tailings dams, such as
valuable minerals, and hazardous substances like heavy metals, overtopping, external erosion, internal erosion, slope instability,
sulphide minerals, and cyanide. An estimate of 14 billion tons of seismic vulnerability, inadequate foundation design, inadequate
tailings were produced in 2010; this waste material is stored in structural design, and mine subsidence.
tailings dams and other storage facilities [15]. An example of a cross
section of a tailings dam is shown in Fig. 1 that illustrates the
important components in determining the condition of a tailings
dam. Tailings dams are constructed in sequential “lifts” over time;
hence, the tailings dam partially depends on the tailings themselves
for support [15]. Usually, a single tailings dam facility is used for
the entire life of the mine, which is approximately 5 to 40 years [18].
Fig. 2. Conceptual framework of the study
Inactive tailings dams are facilities that have been filled or when
tailings production ceases; while active dams are facilities that are
still used to store produced tailings. One of the ways to close a III. METHODOLOGY
tailings dam is by enforcing a wet cover, a common closure strategy,
which requires the surface of tailings dams to be submerged to FMEA is a qualitative risk assessment tool used in
prevent acidic drainage by restricting oxidation of sulphides [18]. mapping out the impact of all harmful events that can occur during
construction or operation of the tailings dam [19]. As shown in Fig.
3, to conduct FMEA, the system must be defined in which the
objective of the project and other pertinent information is
established to set the scope of the study. The first step of FMEA is
to establish the hierarchy of levels. A system is the set of all
components that affect the failure of the structure studied; it is
systematically divided into subsystems down to the basic component
level. The functions of the basic components are then defined, which
Fig. 1. Tailings dam [18]
are used to describe the role of the component in the system. Failure
modes are the description of how failure occurs in a component.
Failure is defined as the cessation of the ability of a component,
96
subsystem, or system to perform one of its functions [19]. Effect is require more stability due to the accumulated wastewater that the
the impact of a failure mode in the performance of the system and dam needs to retain. According to U.S. Environmental Protection
components; furthermore, it consists of outcomes associated with Agency [21], tailings dams that are constructed using upstream
the loss of ability to accomplish a function [19]. The effect on the method have a height increase restriction of 15 to 30 feet per year,
component is the direct effect, while the intermediate effect is the while a height increase that would exceed 50 feet per year are
effect on the subsystem, and the end effect is for the whole system. considered hazardous. Generally, tailings dam heights are kept
The effects are used to formulate measures to prevent failure in each below 200 feet because high embankments, which are greater than
level. Detection measures are methods a failure mode can be 400 feet, pose design and construction problems [21].
detected under normal conditions. The control measures carry out
remedial work to control sequence of its effects by delaying or e) Steepness of tailings dam slope
stopping them; while, mitigation measures reduce end effects and Steep slopes are prone to external erosion and slope
their consequences [19]. instability, which could collapse a portion of the dam embankment
if no protections are in place. Furthermore, steep slopes are more
likely to become unstable during seismic events, such as ground
shaking. Steep-sided sand has a slope of 3H:1V to 4H:1V [21].
f) Drainage system
Tailings are discharged into the dam as slurry, wherein the
solid material separates as it settles into the dam. A suitable drainage
Fig. 3. Flowchart of the summary of procedures system must be used to avoid dam failure due to overtopping in
countries that experience heavy rains [12]. Since the outer slope of
IV. COMPONENTS OF ASSESSING RISK the dam is very sensitive to a high phreatic line, the dam is more
susceptible to failure due to seepage and internal erosion if proper
drainage is not in place.
A. Factors contributing to vulnerability of element at risk
g) Demand of mine products
a) Construction method Tailings dam failures in developed countries happened
The upstream method is the oldest and most commonly before the 21st century; while in developing countries, failures
used construction method. The upstream method starts with a starter happened after the 21st century during their economic growth [12].
dyke in the extreme downstream point then progressively raised Thus, as the mining industry begins to flourish in the period of
towards the upstream side. Upstream constructed dams require economic growth and development, the frequency of failure
fewer construction materials, but the dam has a high phreatic line increases, since developing countries often overlook the proper
and poor stability; furthermore, the dam wall is built on top of design and management of tailings dams when there is an influx of
previously deposited unconsolidated tailings slurry retained behind tailings from mining.
the wall [16]. Thus, the upstream method is more susceptible to
instability from seismic loading compared to other types of h) Date of construction
construction methods, such as downstream and centerline. Most of the tailings dam failures are attributed to tailings
dams that were constructed before the 21st century. According to
b) Deposition of tailings Larrauri [10], prior to the 1970s the construction method used was
The general methods of tailings deposition are single point the upstream method regardless of local condition and there was no
discharge, spigotting, and cycloning. Single point discharge is used real stability analysis. During the 1970s to 1980s a few
to discharge tailings away from the crest; furthermore, the discharge advancements and techniques were utilized to improve the stability
point must be periodically moved to another location to prevent of tailings dams and avoid failure, wherein deterministic approach
unequal raising of the dam section. Spigotting discharges tailings was used for seismic hazards, centerline and downstream method
through small pipes or spigots that originate from multiple points at were used in seismic zones, and usage of safety factors were
regular intervals along the tailings header line [21]. Uniform flow of predominant [10]. During the 1990s to 2000s, seismic stability
tailings, resulting in uniform beaches, are achieved using spigotting. calculations, deterministic and probabilistic assessment of seismic
The distribution characteristics of spigotting is effective in reducing hazards, probable maximum flood determination, seepage
the water level across the dam. Furthermore, uniform beaches are modelling and detection, and limit equilibrium methods for slope
necessary to maintain stability for dams constructed using upstream protection were utilized [10].
method. Cycloning is used to separate the fines from the coarse
fraction in the tailings slurry [21]. i) Tailings dam activity
Active tailings dams are more likely to fail than inactive
c) Embankment material of tailings dam and properties of tailings tailings dams. The increase of water level, wherein the phreatic line
Tailings sands are often used as an inexpensive source of is close to the crest of the downstream slope, is a common cause of
material for construction of tailings dams that can have both inactive dam failures, which leads to slips or overtopping [22].
effective strength and high permeability [21]. Native soil and
tailings sand used as embankment are appropriate for tailings dams
constructed using upstream, downstream, or centerline method. The B. Vulnerability of element at risk
disadvantage of using tailings as embankment include high
susceptibility to piping and high erodible surfaces; moreover, loose a) Overtopping
and saturated tailings are subject to earthquake-induced liquefaction. Overtopping, or the flooding of the top of the tailings dam,
make the slope unstable that causes tailings dam failure [12]. The
d) Tailings dam height initially constructed tailings dam should have a good permeability
The stability analysis carried out for tailings dams under to ensure drainage of water and prevent overtopping. According to
different heights shows that the design height of the tailings dam Lyu et al. [12], the three stages of the tailings dam break process,
should be reduced to ensure safety [12]. A higher tailings dam would which may occur multiple times, are the beginning of destabilization
of tailings dam under adverse condition; interaction of tailings sand
97
with water and formation of debris flow with high energy; and debris g) Inadequate structural design
flow moving downstream. Structural design errors are often caused by failure of
drainage systems, such as the decant facilities. Decant facilities are
b) External erosion used to remove excess water from the tailings dam to manage the
Erosion of tailings dams due to wind and rain can affect water level and phreatic line. Structural failure, such as breakage of
the stability and create environmental problems. Since larger pipes or a destruction of the dam wall caused by the failure of a
particles are near the crest, local small rotational slips may occur component of the decant facility, can cause tailings dam failure.
where the water level meets the downstream slope. Downstream Other forms of surface water control systems aside from decant
slopes suffer slight gullying caused by rainfall; the first sign of a facilities are diversion channels, closed channels, and spillways.
close water level is given by water issuing into one these gullies Spillways are designed as temporary structures constructed with
wherein small slips may develop deepening the gully, which will impervious material to withstand rapid flow velocities [21].
move the surface of that part of the downstream slope further back
into the water level [22]. This leads to larger slips that eat away the h) Mine subsidence
base of the slope until the whole downstream face becomes unstable. Mine subsidence occurs when the tailings dam is built on
The erosion of the embankment can also lead to overtopping [18]. top or near an underground mine [18]. Hence, collapse of the
underground mine can lead to a breach in the foundation, which will
c) Seepage and internal erosion lead to collapse of tailings dam due to the pressure from the flow of
The stability of the tailings dam is greatly influenced by tailings into the underground mine.
the seepage field since tailings dams are permeable [12]. According
to Lyu et al. [12], the factors that increase the susceptibility of
seepage and affect the level of the phreatic line are the level of water C. Hazard occurrence
in the dam, permeability of the initial dam before tailings were Heavy rain, such as typhoons, trigger overtopping and
stored, and rainfall. As the seepage increases, the hydrostatic internal erosion; while, earthquakes trigger seismic instability such
pressure increases, reducing the shear strength of the embankment, as liquefaction of tailings. External erosion, slope instability,
which will cause the failure of the tailings dam. Moreover, structural and foundation failure, and settlement of tailings dams
components of the structure with high permeability are susceptible near underground mines are triggered by rainfall and earthquakes.
to piping or the internal erosion due to seepage [12]. The operational metallic and non-metallic mines were obtained from
the directory of Mines and Geosciences Bureau as of August 2019
d) Slope instability [13, 14]. The occurrence of typhoons and earthquakes in the mines
Rotational sliding results in slope failure ranging from were mapped using Quantum Geographic Information System
local sloughing of tailings at random areas along the face of an (QGIS); the base maps used were from the Typhoon incidence map
embankment to massive circular arc slides that extend over the entire and Earthquake-prone areas map of Center for Environmental
structure [21]. The causes of rotational sliding include changes in Geomatics, & Manila Observatory [3].
water table, changes in permeability of the foundation materials,
disturbance to the embankment caused by impact or vibration, and TABLE I. Operating metallic mines in the Philippines [13]
settlement of foundation materials. Slope instability may also be ID Contractor Location
caused by deformation from static load and partial saturation of Itogon-Suyoc Resources,
areas in the tailings dam designed to remain dry [18]. 1 Sangilo, Itogon, Benguet
Inc.
Benguet Corporation –
e) Seismic vulnerability
2 Acupan Contract Mining Acupan, Itogon, Benguet
Earthquake-induced liquefaction of tailings is the main
Project
cause of damage of high tailings dams, while the influence of
Lepanto Consolidated
seismic inertial force on the seismic stability of the dam is 3 Mankayan, Benguet
Mining Company
secondary. According to Lyu et al. [12], as the height of the dam
4 FCF Minerals Corporation Quezon, Nueva Vizcaya
increases, the seismic performance is reduced. Liquefaction of
tailings are affected by the composition, size, gradation, Filminera Resources
5 Aroroy, Masbate
arrangement, compactness, depth of wetting line, and seismic Corporation
intensity [12]. Studies show that the tailings stored in the dam are 6 Apex Mining Company Inc. Maco, Compostela Valley
saturated and relatively loose [12]. Thus, under seismic loading, the Philsaga Mining Bunawan and Rosario,
7
tailings dam body is susceptible to liquefaction, which makes it Corporation Agusan del Sur
easier for seepage channels to form that can cause collapse. Tmc-Tribal Mining
8 T’boli, South Cotabato
Corporation
f) Inadequate foundation design Alegria, Mainit, Tubod
Greenstone Resources
Weak material beneath the slope of the embankment, low 9 and Bacuag, Surigao Del
Corporation
shear strength of foundation strata, and fissured clay, when not Norte
addressed, are some of the causes of foundation failure [21]. Abra Mining & Industrial
10 Bucay, Abra
Furthermore, the permeability of the foundation affects the stability Corporation
of the embankment; if the foundation material is impervious, the Carmen Copper
11 Biga, Toledo City, Cebu
load of the embankment will generate excess porewater pressure, Corporation
which will reduce the shear strength of the foundation. On the other 12 Philex Mining Corporation Tuba, Benguet
hand, if the foundation material is pervious, excessive seepage can Oceanagold (Philippines),
13 Kasibu, Nueva Vizcaya
lead to piping. Hence, the properties of the tailings beach that will Inc.
become the foundation of the next dyke must comply with the 14 Techiron Resources, Inc. Guiuan, Eastern Samar
requirements for a stable foundation. [21]. Loreto, Dinagat Islands
15 Krominco Inc. (within Parcel III Surigao
Mineral Reservation)
98
TABLE II. Continuation of operating metallic mines [13] TABLE III. Continuation of operating metallic mines [13]
ID Contractor Location ID Contractor Location
Mt. Sinai Mining Oriental Synergy Mining
Homonhon Island, 46 Loreto, Dinagat Islands
16 Exploration and Corporation
Guiuan, Eastern Samar
Development Corporation Citinickel Mines and Narra and Sofronio
47
Rio Tuba Nickel Mining Development Corporation Espanola, Palawan
17 Bataraza, Palawan
Corporation MacArthur and Javier,
48 Nicua Corporation
Tubay, Santiago, and Leyte
18 Agata Processing, Inc.
Jabonga, Agusan del Norte Dulag and Mayorga,
Strongbuilt (Mining)
19 SR Metals, Incorporated Tubay, Agusan del Norte 49 MacArthur, Javier and
Development Corp.
Surigao Integrated Abuyog, Leyte
20 Claver, Surigao del Norte
Resources Corporation 50 Atro Mining-Vitali, Inc. Vitali, Zamboanga City
21 Berong Nickel Corporation Quezon, Palawan
Hinatuan Mining Tagana-an, Surigao del TABLE IV. Operating non-metallic mines in the Philippines [14]
22
Corporation Norte ID Contractor Location
Adnama Mining Resources Apo Land and Quarry
23 Claver, Surigao del Norte 1 Naga, Cebu
Incorporated Corporation
Taganito Mining Republic Cement &
24 Claver, Surigao del Norte 2 Norzagaray, Bulacan
Corporation Building Materials, Inc.
Marcventures Mining and Holcim Mining and
25 Carrascal, Surigao del Sur 3 Norzagaray, Bulacan
Development Development Corporation
CTP Construction and Adlay, Carrascal, Surigao Republic Cement &
26 4 Taysan, Batangas
Mining Corporation del Sur Building Materials, Inc.
CTP Construction and Adlay, Carrascal, Surigao Lazi Bay Resources
27 5 Lazi and Maria, Siquijor
Mining Corporation del Sur Development, Inc.
Carrascal Nickel Republic Cement Iligan City, Lanao del
28 Carrascal, Surigao del Sur 6
Corporation Mindanao, Inc. Norte
Cagdianao Mining Holcim Resources and Iligan City and Lugait,
29 Cagdianao, Dinagat Island 7
Corporation Development Corporation Misamis Oriental
30 Libjo Mining Corporation Libjo, Dinagat Islands Republic Cement Land and Norzagaray and San Jose
8
Oriental Vision Mining Resources, Inc. del Monte, Bulacan
31 Tubajon, Dinagat Island
Philippines Corporation Solid Earth Development
9 San Fernando, Cebu
Loreto and Libjo, Dinagat Corp.
32 Century Peak Corporation
Islands Holcim Mining and
Bunawan District, Davao
Benguetcorp Nickel Mines 10 Development Corporation –
33 Sta. Cruz, Zambales City
Inc. HMDC
34 Eramen Minerals, Inc Sta. Cruz, Zambales Republic Cement Iligan, Kiwalan and Iligan City,
11
Zambales Diversified Sta. Cruz and Candelaria, Inc. Lanao del Norte
35
Metals Corporation Zambales Maria Cristina Chemical
LNL Archipelago Minerals, 12 Industries (MCCI) Kiwalan, Iligan City
36 Sta. Cruz, Zambales Corporation
Inc.
Claver, Surigao del Norte Northern Cement
13 Sison, Pangasinan
(Within Parcel I of Corporation
37 Claver Mineral Dev't. Corp.
Surigao Mineral 14 Teresa Marble Corporation Antipolo City, Rizal
Reservation) Island Quarry and
15 Antipolo City, Rizal
Sinosteel Phils. H. Y. Aggregates Corp.
38 Loreto, Dinagat Islands
Mining Corporation Republic Cement &
16 Teresa, Rizal
Libjo and Tubajon, Building Materials, Inc.
39 Wellex Mining Corporation
Dinagat Islands Holcim Mining and
Pacific Nickel Philippines, Cagdianao, Surigao Del Development Corporation Doña Remedios Trinidad
40 17
Inc Norte and Doric Marble and Norzagaray, Bulacan
Dinapigue Mining Corporation
41 Dinapigue, Isabela
Corporation Ibalong Resources and Camalig and Guinobatan,
18
Basilisa and San Jose, Development. Corporation Albay
AAM-Phil Natural
Dinagat Islands (Parcel II Bohol Limestone
42 Resources Exploration and 19 Garcia Hernandez, Bohol
of Surigao Mineral Corporation
Development Corporation
Reservation) Solid North Mineral
20 San Ildefonso, Bulacan
Basilisa and San Jose, Corporation
AAM-Phil Natural
Dinagat Islands (Parcel II 21 Eagle Cement Corporation San Ildefonso, Bulacan
43 Resources Exploration and
of Surigao Mineral Bigte and Norzagaray,
Development Corporation 22 Spar Development Co., Inc.
Reservation) Bulacan
Mil-Oro Mining Mati and Gov. Generoso, Solid Earth Development
44 23 San Fernando, Cebu
Corporation Davao Oriental Corporation
Panamaon, Loreto, Rio Tuba Nickel Mining
45 Century Peak Corporation 24 Bataraza, Palawan
Dinagat Island Corporation
99
TABLE V. Continuation of operating non-metallic mines [14] upon by extreme weather events, such as typhoons. Using the spot
ID Contractor Location map in Fig. 6 and Fig. 7, it was determined that mines located in
Holcim Mining and South Cotabato, Surigao Del Norte, Dinagat Island, Agusan del
Bacnotan and Balaoan, La Norte, and Sarangani experience approximately 48 to 75 hits per
25 Development Corporation-
Union year, while mines in Agusan del Sur, Surigao del Norte, and Surigao
HMDC
Holcim Mining and del Sur experience 76 to 100 earthquake hits per year. Thus, tailings
Davao City and Maasim, dams near the mines located in these areas are susceptible to failure
26 Development Corporation-
Sarangani triggered by earthquake events. In addition, mines located in
HMDC
Holcim Resources and Iligan City and Lugait, Benguet, Nueva Vizcaya, Pangasinan, La Union, and Bataan has a
27 high or very high typhoon incidence, and experience 48 to 100
Development Corporation Misamis Oriental
Holcim Mining and earthquake hits per year; thus, tailings dam near the mines in these
28 Development Corporation- Bacnotan, La Union areas are susceptible to failure triggered by earthquakes and rainfall.
HMDC
29 Teresa Marble Corporation Antipolo City, Rizal
30 Teresa Marble Corporation Antipolo City, Rizal
Naga and Pinamungahan,
31 Quarry Ventures Phils., Inc.
Cebu and Toledo City
Rapid City Realty and
32 Teresa, Rizal
Development Corp.
Rapid City Realty and
33 Antipolo City, Rizal
Development Corporation
34 Apo Land and Quarry Corp. Naga, Cebu
Concrete Aggregates
35 Angono, Rizal
Corporation
Concrete Aggregates
36 Angono, Rizal
Corporation
Rapid City Realty and
37 Antipolo City, Rizal
Development Corporation
Asensio Pinzon Aggregate
38 Rodriguez, Rizal
Corp.
Rapid City Realty and
39 Antipolo City, Rizal
Development Corporation
Rapid City Realty and
40 Antipolo City, Rizal
Development Corporation
Island Quarry and
41 Antipolo City, Rizal
Aggregates Corporation
Rapid City Realty and
42 Antipolo City, Rizal
Development Corporation
Fig. 4. Typhoon incidence in operating metallic mines
Rapid City Realty and
43 Baras, Rizal
Development Corporation
JLR Construction and Naga and Minglanilla,
44
Aggregates, Inc. Cebu
45 Hardrock Aggregates, Inc. Antipolo City, Rizal
Rolando B. Gimeno and La
Concepcion Construction
46 Antipolo City, Rizal
and Development
Corporation
Gozon Development
47 Antipolo City, Rizal
Corporation
Dolomite Mining Alcoy and Dalaguete,
48
Corporation Cebu
UBS Marketing
49 Legazpi City, Albay
Corporation
50 Heirs of Elias E. Olegario Mangatarem, Pangasinan
51 Orophil Stonecraft, Inc. Baao, Camarines Sur
52 Rockmix, Inc. Balanga, Bataan
Island Quarry and Palangoy, Binangonan,
53
Aggregates Corporation. Rizal
100
TABLE VII. Typhoon incidence in operating non-metallic mines
Typhoon
ID Location
Incidence
2 Norzagaray, Bulacan High
3 Norzagaray, Bulacan High
4 Taysan, Batangas High
8 San Jose del Monte, Bulacan High
14 Antipolo City, Rizal High
15 Antipolo City, Rizal High
16 Teresa, Rizal High
Doña Remedios Trinidad and
17 High
Norzagaray, Bulacan
18 Camalig and Guinobatan, Albay High
20 San Ildefonso, Bulacan High
21 San Ildefonso, Bulacan High
22 Bigte and Norzagaray, Bulacan High
29 Antipolo City, Rizal High
30 Antipolo City, Rizal High
32 Teresa, Rizal High
33 Antipolo City, Rizal High
35 Angono, Rizal High
36 Angono, Rizal High
37 Antipolo City, Rizal High
38 Rodriguez, Rizal High
39 Antipolo City, Rizal High
Fig. 6. Earthquake-prone operating metallic mines
40 Antipolo City, Rizal High
41 Antipolo City, Rizal High
42 Antipolo City, Rizal High
43 Baras, Rizal High
45 Antipolo City, Rizal High
46 Antipolo City, Rizal High
47 Antipolo City, Rizal High
49 Legazpi City, Albay High
51 Baao, Camarines Sur High
53 Binangonan, Rizal High
101
TABLE X. Typhoon incidence in earthquake-prone metallic mines TABLE XII. Tailings dam failure in the Philippines
Typhoon Earthquake ID Date Mine/Project and Location
ID Location
Incidence Hits A 28/Oct/16 Antamok, Baguio (Benguet Corp.)
1 Sangilo, Itogon, Benguet High 76 to 100 B 2/Aug/12 Padcal No 3, Benguet (Philex)
2 Acupan, Itogon, Benguet High 76 to 100 Bayarong dam, San Marcelino, Zambales
C 11/Sep/02
3 Mankayan, Benguet Very High 48 to 75 (Benguet Corp-Dizon Copper-Silver Mines)
4 Quezon, Nueva Vizcaya High 48 to 75 Camalca dam, San Marcelino, Zambales
D 27/Aug/02
13 Kasibu, Nueva Vizcaya High 48 to 75 (Benguet Corp-Dizon Copper-Silver Mines)
12 Tuba, Benguet High 76 to 100 E 9/Aug/99 Toledo City (Atlas Con Mining Corp)
TP7, Placer, Surigao Del Norte (Manila
F 26/Apr/99
TABLE XI. Typhoon incidence in earthquake-prone non-metallic Mining Corp)
mines Sibutad Gold Project, Zamboanga Del Norte
G 27/Jun/98
Typhoon Earthquake (Philex Mining Corp)
ID Location
Incidence Hits Sibutad Gold Project, Zamboanga Del Norte
H 6/Nov/97
13 Sison, Pangasinan High 76 to 100 (Philex Mining Corp)
Bacnotan and Balaoan, La Marcopper Mine,Tapian Pit, Marinduque
25 Very High 76 to 100 I 24/Mar/96
Union Island (Placer Dome & President Marcos)
28 Bacnotan, La Union Very High 76 to 100 Bulawan Mine, Sipalay River, Negros
J 8/Dec/95
50 Mangatarem, Pangasinan High 48 to 75 Occidental (Philex Mining Corp)
52 Balanga, Bataan High 48 to 75 Bulawan Mine, Sipalay River, Negros
K Apr/95
Occidental (Philex Mining Corp)
V. TAILINGS DAM FAILURE IN THE TP5, Placer, Surigao del Norte (Manila
L 2/Sep/95
Mining Corp) 2nd event
PHILIPPINES Marcopper, Mogpog, Marinduque Island
M 6/Dec/93
(Placer Dome-President Marcos)
A. Local tailings dam failure Baguio gold district, Itogon-Suyoc (Benguet
N 26/Jun/93
Corp)
O 2/Jan/92 TP2, Tailings Pond, Tubu, Benguet
Placer, Surigao Del Norte (Manila Mining
P 9/Jul/87
Corp)
TP3, Mankayan District (Benguet Corp
Q 17/Oct/86
subsidiary Lepanto Con Mining Co)
R 8/Nov/82 TP3, Sipalay (Maricalum Mining Corp)
102
TABLE XV. Continuation of tailings dam failure in the Philippines Heavy rain or typhoons trigger 56% of the recorded
ID Natural Hazard Cause Release (cu.m.) tailings dam failures listed in TABLE XIV and TABLE XV; in
G Undetermined OT Undetermined addition, the majority of tailings dam failures were caused by
H Heavy rain OT Undetermined overtopping. On the other hand, 33% of tailings dam failures are
I Earthquake ST 1,600,000 caused by structural failure, which include the failure of drainage
J Undetermined ST Undetermined systems. During an intense rainfall, the increase in volume of water
K Heavy rain U 50,000 may exceed the capacity of the drainage system which may lead to
L Heavy rain FN 50,000 failure or a breach in pipes. Failure of the foundation to support the
M Undetermined U Undetermined load of the tailings dam, is also 17% of the cause of tailings dam
failure, wherein the excess volume brought by heavy rain may
N Undetermined OT Undetermined
exceed the designed load that the structure is able to support. Since
O Earthquake FN 80,000,000
major weather events are more frequent than major seismic
P Undetermined U Undetermined
activities, only 11% of the failed tailings dams in the Philippines
Q Heavy rain ST 100,000 were triggered by earthquakes, wherein the cause of failures were
R Heavy rain FN 25,000,000 largely attributed to structural failure and foundation failure.
Earthquake events may cause a rupture in the pipes underground
TABLE XVI. Continuation of tailings dam failure in the Philippines designed for the drainage system. Furthermore, seismic events may
ID Effects of Disaster on Elements Source trigger foundation failure by rupturing the foundation strata.
Contaminated Liang River, Ambalanga
A [24]
River, and Agno River Surface water was mostly affected by the tailings dam
Contaminated Balog River, Agno River, failure, while a few reports indicate the effects of acid mine
B [1, 24]
and San Roque Dam drainage. Nearby croplands were also affected by the sediments
Contaminated Mapanuepe Lake and from the tailings dam that contaminated fertile croplands by burying
C [23, 24]
Santo Tomas River it. Most of the tailings that were stored in the failed tailings dam
Contaminated Mapanuepe Lake and were from the production of gold and copper, which usually contains
D [23, 24]
Santo Tomas River hazardous substances, such as heavy metal waste and cyanide waste
E Contaminated Sapangdaku River [9, 24] [18]. From the reports, the tailings spill is composed of acidic mine
Contaminated 51 hectare of cropland and [2, 8, 20, 23, water and sediments that contain hazardous substances; thus, bodies
F
Batayakan area 24] of water, such as Boac River and Mogpog River were declared
G Undetermined [24] biologically dead after the tailings spill.
H Fish kills in nearby bodies of water [20, 24]
Contaminated Makulaquit River and Boac
River System, and Barangay Hinapulan B. Padcal Mine Site Tailings Pond 3 in Padcal, Tuba, Benguet
[4, 6, 11, 17,
I cropland; Boac River Biologically Dead; Tailings Pond 3 was used to store tailings from Padcal
24]
Seepage in Hinapulan creek; Acid Rock Mine Site, wherein the main products obtained are gold and copper.
Drainage in Groundwater According to AGHAM [1], Padcal Mine pioneered the underground
J Contaminated Sipalay River [20, 24] block cave; a type of mining that creates a cave and extracts a large
K Coastal Area Pollution in Placer Bay [23, 24] volume of rock from the formation of surface depression. Tailings
L Coastal Area Pollution in Placer Bay [8, 20, 24] Pond 3 has a catchment area of 350 hectares wherein the south end
Contaminated Mogpog Town and of the main dyke consists of two main penstock, A and B. Tailings
Mogpog River; Mogpog River [6, 7, 17, 24] Pond 3 started construction in 1983 and commissioned in 1992, with
M
Biologically Dead; Acid Rock Drainage a lifespan of 18 to 20 years [1]. Moreover, the dam crest went
in Groundwater beyond the permitted elevation of 600 meters above sea level. Also,
N Undetermined [24] fault lines, such as Albian and Santa Fe, traverse Tailings Pond 3.
Heavy siltation in Irrigation systems
O [20, 23, 24]
downstream
P Fish kills in Placer, Surigao del Norte [20, 24]
Q Contaminated Abra River [20, 24]
Laid 1.5 meters thick of sediment on
R [23, 24]
agricultural land
103
Penstock A was breached that resulted in the discharge of TABLE XX. Risk scoring
tailings, which contaminated Balog River, Agno River, and San Risk class Description Risk Number (RN)
Roque Dam, and polluted nearby fertile lands and vegetation. The I Low 1 to 15
incident was triggered by heavy rain brought by typhoon Ferdie and II Medium 16 to 30
Gener. There were five major incidents of discharge reported on III High 31 to 45
August 1, 4, 11, and 30, and September 13, 2012. Fig. 11 shows the
idealized cross section of Tailings Pond 3 prior to the incident and
Fig. 12 shows the cross section after the incident, both illustrations
VII. QUALITATIVE RISK ASSESSMENT
from Mines and Geoscience Bureau. Sinkholes, due to the OF PADCAL MINE SITE TAILINGS
underground block cave, were also found in the tailings pond, which POND 3
was plugged using concrete spheres following the discharge to also
prevent tailings from leaking out [5]. The tunnels were sealed by
concrete to prevent leakage; thus, spillways were built to drain water A. Preliminary qualitative checklist rating
from the tailings dam instead of using penstocks and underground
drain tunnels, since the Department of Environment and Natural Padcal Mine Site Tailings Pond 3 is located in Tuba,
Benguet which is an earthquake-prone area, hit by approximately 76
Resources approved the life extension of Tailings Pond 3 to store
to 100 hits per year, and has a very high typhoon incidence. TABLE
waste from the operating Padcal Mine Site [5].
XXI shows the preliminary qualitative risk assessment for Padcal
Mine Site Tailings Pond 3. Each vulnerability of the tailings dam
VI. PRELIMINARY QUALITATIVE was evaluated by multiplying the corresponding hazard and
CHECKLIST RATING probability of the hazard. Padcal Mine Site Tailings Pond 3 is
classified a Class III risk since the likelihood that a tailings dam
TABLE XVII, TABLE XVIII, and TABLE XIX shows breach may occur due to overtopping, inadequate foundation design,
the rating of the vulnerabilities, hazards, and hazard occurrence for and inadequate structural design is high. Thus, further analysis to
tailings dams in the Philippines. The ratings are based on the studies determine which component failure will lead to tailings dam breach
and literature that were reviewed, which showed the critical factors is necessary to properly assess the risk.
that affect tailings dam failure in the local and global setting.
Equation 1 is used to obtain the risk class of a tailings dam to TABLE XXI. Padcal Mine Site Tailings Pond 3 risk class
determine which tailings dam should have a focused risk reduction Risk number computation Risk class
measure. Since tailings dam failures in the Philippines are mostly V1 × H3 × P5 45 III (High)
caused by typhoons, typhoons were given a high hazard rating. V2 × H3 × P5 15 I (Low)
Moreover, the common vulnerability of tailings dam failure in the V3 × H3 × P5 15 I (Low)
Philippines are overtopping, inadequate foundation design, and V4 × H3 × P5 30 II (Medium)
inadequate structural design; thus, the said vulnerabilities were V5 × H2 × P5 20 II (Medium)
assigned a rating of 3. V6 × H3 × P5 45 III (High)
V7 × H3 × P5 45 III (High)
Equation 1 V8 × H3 × P5 30 II (Medium)
104
TABLE XXIII. Failure modes of components and its root causes TABLE XXIV. Continuation of failure modes of components and
Component Failure Mode Root Causes its root causes
Heavy rainfall; excessive Component Failure Mode Root Causes
External
wind action; weathering; Changes in water level;
Erosion (EE)
vibration changes in permeability;
Slope Instability
Changes in water level; disturbance of due to
(SLE)
changes in permeability; vibration; deformation due
Mountain
disturbance due to vibration; to static load
Slope
Slope Instability settlement of foundation; Seismic activity; mechanical
Seismic
(SLI) partial saturation of areas properties of mountain slope
Vulnerability
designed to remain dry; material; uncompacted
Main Dyke (SV)
deformation due to static mountain slope soil particles
load
Seismic Seismic activity; low shear TABLE XXV. Effects analysis for Tailings Pond 3
Vulnerability strength of embankment and Failure Intermediate
Component Direct Effect End Effect
(SV) tailings Mode Effect
Heavy rainfall; inadequacy Collapse at Reduction
Deformation
Overtopping of drainage system; increase downstream in stability
SV of
(OT) in sediments due to eroded face of of tailings
embankment
materials tailings dam dam
Heavy rainfall; excessive Local Eroded
External
wind action; weathering; rotational materials at Reduction
Erosion (EE) EE
vibration slips on the the bottom of of dam
Seepage and Permeable embankment; rise crest the dyke height that
Main Dyke
Internal Erosion of water table and phreatic Circular arc Collapsed prevents
(SIE) line slide at materials at overflow
SLE
Offset Dyke Seismic Seismic activity; low shear downstream the bottom of of tailings
Vulnerability strength of embankment and face the dyke
(SV) tailings Overflow of
Destabilizatio Tailings
Heavy rainfall; inadequacy tailings over
OT n of dyke dam
Overtopping of drainage system; increase crest of main
system breach
(OT) in sediments due to eroded dyke
materials Eroded
Inadequate material
Penstock A Seismic activity; external
Structural Local will
and B forces from tailings Reduction in
Design (ST) rotational increase
EE compactness
Drain Inadequate Seismic activity; external slips near volume
and stability
Tunnel A Structural forces from tailings and water level retained in
and B Design (ST) embankment tailings
Inadequate Seismic activity; increase dam
Spillway Structural water flow from extreme Weaker
Design (ST) weather events Seepage of dam wall
Inadequate Weak material beneath slope tailings Piping or due to
Foundation of embankment; low shear SIE water; internal voids or
Design (FN) strength; fissured clay layer reduction of erosion presence
Offset
Seepage and Permeable foundation shear strength of seepage
Dyke
Internal Erosion material; rise of water table channels
(SIE) and phreatic line Collapsed
Seismic activity; static load material
Foundation Deformation; will
Mine from weight of tailings; rise
Strata liquefaction Collapse at increase
subsidence of water table and phreatic SV
(MS) line, which also increases the of tailings upstream face volume
load sand retained in
Seismic activity; low shear tailings
Seismic dam
strength of foundation
Vulnerability Overflow of Mobilization Offset
material and tailings;
(SV) tailings over of tailings dyke is
uncompacted tailings OT
Mountain External Heavy rainfall; excessive crest of offset onto main unable to
Slope Erosion (EE) wind action; weathering; dyke dyke function
vibration Damage on
Penstock A Leakage of
Seepage and Permeable mountain slope ST the concrete
and B tailings on Tailings
Internal Erosion soil particles; rise of water plugs
mountain dam
(SIE) table and phreatic line Drain Damage on
slope and breach
Tunnel A ST the concrete
ground
and B plugs
105
TABLE XXVI. Continuation of effects analysis for Tailings Pond 3 The structural component of the drainage system must be
Failure Intermediate protected because it is prone to failure that will likely lead to tailings
Component Direct Effect End Effect
Mode Effect spill. In TABLE XXVI, it can be observed that the end effect of the
Overflow failure mode for penstocks and drain tunnels is tailings dam breach.
Inability to Inadequacy to of tailings Based on the incident of Tapian Pit in Marinduque, Philippines in
regulate flow remove dam; March 1996, discontinuities in the structure, such as plugged tunnels
Spillway ST are susceptible to failure since the pressure from the tailings is
of water; excess water tailings
damage on dam dam considerably large [18]. In addition, a breach already occurred in
breach penstock A that mobilized a large volume of tailings into Balog
Reduction of River; thus, the plug in the drain tunnels must be effective and
Excess regularly monitored to assure that the same penstock will not be
shear Tailings
porewater breached again. In addition, the end effect of the failure for spillways
FN strength; dam
pressure; is also tailings dam breach because spillways are temporary
internal breach
seepage structures that are removed before the tailings dam is raised and will
erosion
Weaker be reconstructed again after.
Seepage of dam wall
Instability; Failure of tailings dam foundation due to mine subsidence,
tailings due to
internal seismic vulnerability, and inadequate foundation design will likely
SIE water; voids or
Foundation erosion or lead to mobilization of a large volume of tailings. The Padcal Mine
reduction of presence
Strata piping Site pioneered underground block cave mining, in which the
shear strength of seepage
channels incidence of tailings dam breach in 2012 exposed sinkholes due to
Tailings underground mining, which was then filled using concrete spheres.
MS Settlement Instability dam However, sinkholes may still be present near the tailings dam, which
breach will increase the likelihood of settlement since the dam is still active
Settlement; and more tailings are acquired that will increase the load that the
Tailings foundation must support. Furthermore, the location of Padcal Mine
liquefaction Permanent
SV dam Site Tailings Pond 3, is prone to earthquakes; thus, the likelihood of
of tailings deformation
breach settlement triggered by earthquake is also significant. Since the
sand; rupture
Local Eroded tailings dam has an existing plugged underground drain tunnel
Reduction in illustrated in Fig. 11 and Fig. 12, the likelihood of foundation failure
rotational and
EE compactness may increase due to the discontinuity of the component. Fissures in
slips near collapsed
and stability the foundation strata may be present near the drain tunnels, which
water level material
Deformation Collapse at will may exacerbate tailings dam failure if not regularly monitored.
SV
of slope upstream face increase
volume VIII. CONCLUSION
Circular arc Reduction in
retained in
Mountain SLE slide at compactness
tailings Storage facilities, such as tailings dams, must be protected
Slope upstream face and stability
dam to avoid failure that will lead to discharge of a large volume of
Weaker tailings that will directly affect the groundwater, surface water, soil,
Seepage of dam wall and vegetation. Thus, qualitative risk assessment, such as FMEA, is
Instability;
tailings due to conducted to mitigate said disaster. To conduct the risk assessment,
internal
SIE water; voids or the vulnerability of element at risk and hazard occurrence must be
erosion or
reduction of presence identified. Since there are no available directories of active and
piping effect
shear strength of seepage inactive tailings dams, the hazard occurrence for the mines was
channels determined with the assumption that its designated tailings dam is
nearby. Mines in Zambales, Dinagat Island, Leyte, Bulacan,
The effects analysis of Padcal Mine Site Tailings Pond 3 Batangas, Rizal, Albay, and Camarines Sur have high typhoon
shows that overtopping will likely lead to tailings dam spillage incidence, while mines in Abra and Isabela have very high typhoon
among any other failure modes for the main dyke. Overtopping is incidence. On the other hand, mines in South Cotabato, Surigao Del
mostly caused by heavy rainfall and mismanagement of water level. Norte, Dinagat Island, Agusan del Norte, and Sarangani experience
Since there are several strong typhoons in the Philippines, the 48 to 75 hits per year, while mines in Agusan del Sur, Surigao del
likelihood of rain-induced failure is high as observed from past Norte, and Surigao del Sur experience 76 to 100 earthquake hits per
tailings dam incidents in the country. Furthermore, once the tailings year. Furthermore, Mines in Benguet, Nueva Vizcaya, Pangasinan,
overflow from the dam wall, it will likely lead to tailings spill due La Union, and Bataan have a high or very high typhoon incidence,
to the high energy flow of debris that can also mobilize part of the and experience 48 to 100 earthquake hits per year; thus, risk
embankment, which will lead to the collapse of the entire tailings assessment should be conducted.
dam wall. Overtopping in the offset dyke will not immediately lead
to the collapse of the dam wall since the main dyke can still prevent Since FMEA is an intensive analysis that requires the
the tailings from spilling onto the environment; however, once the investigation of every component of the tailings dam, a preliminary
offset dyke is unable to function, the embankment material of the qualitative checklist rating was conducted in which the factors that
offset dyke will serve as an additional external force that the main increased the risk of tailings dam breach were evaluated and
dyke has to resist aside from the external force from the flow of categorized as low, medium, or high risk. In applying FMEA, the
debris and tailings water. Hence, external erosion, slope instability, factors that contribute to the vulnerability of the element at risk is
and seismic vulnerability of the main dyke must be avoided to ensure related to the root cause of failure modes; on the other hand, the
that it is strong enough to resist external forces from embankment vulnerability of the element at risk are related to the failure modes.
material of offset dyke and tailings. The factors that contribute to the vulnerability of the tailings dams
are the construction method, deposition of tailings, embankment
106
material of tailings dam and properties of tailings, tailings dam [6] Dizon, N., “The Marcopper mine spill and the unending wait
height, steepness of tailings dam slope, drainage system, demand of for justice,” retrieved from
mining products, date of construction, and tailings dam activity. The https://www.philstar.com/headlines/2019/04/11/1909180/mar
factor that contributes the most to the vulnerability is the copper-mine-spill-and-unending-wait-justice, 2019.
construction method; using an upstream method increases the [7] Dizon, N., and Vera Files, “The Marcopper disaster: A tragedy
likelihood of failure caused by its poor stability and high phreatic that continues in people's veins,” retrieved from
line. The vulnerability of the tailings dams are overtopping, seepage https://news.abs-cbn.com/spotlight/04/04/19/the-marcopper-
and internal erosion, external erosion, slope instability, seismic disaster-a-tragedy-that-continues-in-peoples-veins, 2019.
vulnerability, inadequate foundation design, inadequate structural [8] Environmental Science for Social Change, “Manila Mining
design, and mine subsidence. Since Padcal Mine Site Tailings Pond Corporation in the Surigao del Norte Mining District,”
3 was classified as a Class III risk, it was further analyzed using Philippine Working Group on Mining, 2008.
FMEA. The effects analysis of Padcal Mine Site Tailings Pond 3 [9] Landingin, R., and Aguilar, J., “Dirty Past,” retrieved from
shows that overtopping, inadequate foundation design, inadequate https://www.rappler.com/business/special-
structural design, and mine subsidence will cause tailings dam report/whymining/whymining-latest-stories/7195-dirty-past,
breach. Furthermore, the discontinuities due to underground 2008.
structures will lead to tailings dam breach, which is exacerbated by [10] Larrauri, P., “Assessing Risks of Mine Tailing Dam Failures,”
the additional forces from rainfall and earthquake. Norges Bank Investment Management, 2017.
[11] “Latest mining disaster further signal to junk mining law,”
IX. RECOMMENDATION retrieved from https://www.ibon.org/latest-mining-disaster-
further-signal-to-junk-mining-law/, 2015.
The accuracy of the rapid risk assessment using the [12] Lyu, Z., Chai, J., Xu, Z., Qin, Y., and Cao, J., “A
preliminary qualitative risk checklist may be improved by comprehensive review on reasons for tailings dam failures
incorporating weights for each factor based on quantitative data that based on case history,” Advances in Civil Engineering, 2019.
may be obtained once the tailings dams are monitored using [13] Mines and Geosciences Bureau, “Directory of Metallic
measurement devices such as piezometers and sensors to monitor Operating Mines in the Philippines,” 2019.
the condition of the tailings dam. Since data is not sufficient to [14] Mines and Geosciences Bureau, “Directory of Non-Metallic
conduct a more detailed FMEA, the FMEA of the Padcal Mine Site Operating Mines in the Philippines,” 2019.
Tailings Pond 3 may still be improved with additional data. FMEA [15] Owen, J. R., Kemp, D., Lebre, E., Svobodova, K., and Perez
for tailings dams in other sites in the Philippines can be conducted Murillo, G., “Catastrophic tailings dam failures and disaster
properly with sufficient data to analyze the failure modes based on risk disclosure,” International Journal of Disaster Risk
the actual design, which can be compared to the industry standard. Reduction, 2020.
Detection, control, and mitigation measures may also be formulated [16] Penman, A. M., Williamson, J. G., Phillips, J., Pimenta de
with the proper data from monitoring instruments in the dam so that Avila, j, Abadjiev, C. B., Klohn, E. J., … Noguera, G.,
the impending disaster can be prevented using the detection “Tailings Dams Risk of Dangerous Occurrences,” Commission
measures and controlled using the control measures. In addition, the Internationale des Grands Barrages, 2001.
relation between failure modes should be considered in the effects [17] Plumlee, G. S., Morton, R. A., Boyle, T. P., Medlin, J. H., and
analysis to provide a more accurate qualitative risk assessment that Centeno, J. A., “An Overview of Mining-Related
can also be used as a basis for quantitative risk assessment. Environmental and Human Health Issues,” Marinduque Island,
Philippines. U. S. Geological Survey, 2000.
ACKNOWLEDGEMENT [18] Roche, C., Thygesen, K., and Baker, E., “Mine Tailings
Storage: Safety Is No Accident,” Nairobi and Arendal: United
Engr. Cielo Frianeza wishes to thank Dr. Andres Oreta for Nations Environment Programme and GRID-Arendal, 2017.
the guidance with regards to disaster risk reduction. [19] Santos, R., Caldeira, L., and Serra, J., “FMEA of a tailings
dam,” Georisk: Assessment and Management of Risk for
REFERENCES Engineered Systems and Geohazards, 2012.
[1] AGHAM, “Environmental Investigation Mission on the [20] Stark, J., Li, J., and Terasawa, K., “Environmental Safeguards
Impacts of the Philex Mining Corporation (PMC) Mine and Community Benefits in Mining: Recent Lessons from the
Tailings Pond 3 Failure,” 2013. Philippines,” In Environmental Safeguards and Community
[2] Carating, R. B., and Galanta, R. G., “The Soils of the Benefits in Mining: Recent Lessons from the Philippines, 2006.
Philippines,” Springer Science Business Media Dordrecht [21] U.S. Environmental Protection Agency, “Design and
Heidelberg, 2014. Evaluation of Tailings Dams,” Washington, DC: Office of
[3] Center for Environmental Geomatics, & Manila Observatory, Solid Waste Special Waste Branch, 1994.
Mapping Philippine Vulnerability to Environmental Disasters, [22] Wills, B. A., and Finch, J. A., “Tailings Disposal” in Wills'
2005. Mineral Processing Technology: An Introduction to the
[4] de la Cruz, G., “Look Back: The 1996 Marcopper mining Practical Aspects of Ore Treatment and Mineral Recovery (8th
disaster,” retrieved from https://www.rappler.com/move- ed., pp. 439–448), Waltham, MA: Butterworth-Heinemann,
ph/issues/disasters/165051-look-back-1996-marcopper- 2016.
mining-disaster, 2017. [23] Wood, H., “Disasters and Minewater,” London: IWA
[5] Dinglasan, R. R., “No stranger to mine spills, Philex still wants Publishing, 2012.
rehab of aging tailings pond,” retrieved from [24] “World Mine Tailings Failures from 1915”, Retrieved from
https://www.gmanetwork.com/news/news/specialreports/2940 https://worldminetailingsfailures.org/, 2019.
83/no-stranger-to-mine-spills-philex-still-wants-rehab-of-
aging-tailings-pond/story/, 2013.
107
Information and
Communications
Technology
Session Chair : Dr. Edwin Sybingco
Moderator : Engr. Jose Martin Maningo
Panelists : Dr. Kathleen Aviso
Dr. Angela Bandala
Engr. John Anthony Jose
Predicting Stag and Hare Hunting Behaviors Using
Hidden Markov Model
Rex Bringula1,a* and Ma. Mercedes T. Rodrigo2,b
1,2
Ateneo de Manila University
a
rexbringula@gmail.com, bmrodrigo@ateneo.edu
Abstract— In this paper, we used Hidden Markov Model (HMM) would be awarded to three groups of students with the highest
to describe the gaming behaviors of students and whether they will game scores. Twenty-five boys and 12 girls with an average of 11
exhibit “stag” or “hare” hunting behavior in a mobile game for years old participated in the study. They had varying degrees of
mathematics learning. We found that there is a 99% probability mathematics abilities – eight were low-performing, 13 were
that the students will stay either as stag or hare hunters. Our results average-performing, and 16 were high-performing.
also suggest that they would choose arithmetic problems involving
addition. These game behaviors are not beneficial to learning Speed is a game setting that determines the pace of the
because they are only exhibiting mathematical skills they already game. Students may choose a very slow (2 points), slow (5 points),
know. The results of the study show that stag and hare hunters medium (10 points), fast (15 points), or very fast (20 points) mode.
have unique traits that separate the one from the other. Students were informed that there were no deductions of points for
wrong answers but on the time limit. Students that chose the first-
Keywords—collaborative learning, game behavior, three settings were labeled as hare hunters; otherwise, they were
hidden Markov model, Viterbi algorithm labeled as stag hunters. We used the hidden Markov model (HMM)
to predict whether students would exhibit a stag or hunting
I. INTRODUCTION behavior given the sequence of difficulty level and type of problem
solved. These variables were used because we found in our
A collaborative learning environment posits that students separate study that these were the significant variables of a
learn while they are in a social group [4]. As social learners, decision tree model that could classify stag and hare hunter
students may display varying degrees of participation [1]. Stag and behaviors [3]. There are 20 possible game settings (or states) (5
hare hunting behaviors are the tendency of a leaner to contribute to difficulty problems times 4 types of problem) that a participant
the scores of the group by choosing either a high-risk mode but may choose from. Furthermore, there are 400 possible transitional
with higher points or a low-risk mode but with lesser points [2]. states (20 states multiplied by 20 states). Initial probabilities for the
This game behavior is relatively unknown in the field of a HMM were determined based on the log files. HMM using the
collaborative mobile learning environment. In this study, we Viterbi algorithm was implemented in Jupyter Notebook. Thirty
attempted to describe the game behaviors of the students and random samples with ten random observations were fed into the
predict whether students will exhibit the stag and hunting HMM model to determine the possible game behavior of the
behaviors. The output of the study will serve as a basis in the students. The random samples were generated through a
development of adapted mobile-based learning for mathematics. spreadsheet. The number of random observations is based on the
average number of questions a student can solve per game session.
On average, one student solved nine questions per game session.
II. METHODOLOGY
This study used the dataset collected by Bringula and III. RESULTS AND DISCUSSION
Rodrigo [2]. The dataset is composed of interaction log files that
were generated through the use of a mobile collaborative game We found that there is a very high probability (99%) that
named Ibigkas!Math. It is a mobile-based learning application for the students will stay at their current game behavior (Figure 1). The
grades 1 to 6 students. It is a collaborative game that covers result suggests that the students will persistently choose a game
arithmetic problems (addition, subtraction, multiplication, and mode based on the speed setting. This means that they will choose
division of whole numbers and fractions). The students utilized the a game setting that they are comfortable with and are less likely to
mobile game for 15 minutes in an experiment conducted during explore new game speed settings. This game behavior may not be
their class session. The application generates arithmetic problems beneficial to mathematics learning since students are only
and it displays on one of the team members' mobile devices. The exhibiting the skills that they already know.
player has to read aloud the arithmetic problems. The answers are
presented in multiple-choice. The correct answer will appear on The graph in Fig 1 also shows that they choose only 10
one of the team members' devices. The complete description of this states out of the 20 possible states. These states are easy and
game is discussed extensively in Bringula and Rodrigo [2]. addition (eaadd), easy and multiplication (eamul), easy and
division (eadiv), very easy and subtraction (vesub), very easy and
The dataset contains 4,628 solved problems. The log files addition (veadd), very hard and addition (vhadd), medium and
have six features, such as difficulty level, type of problem solved, addition (mdadd), medium and multiplication (mdmul), medium
speed, time spent, and the number of correct attempts. The and subtraction (mdsub), and hard and addition (haadd). It is worth
difficulty level and type of problem solved are game modes. In the noting that they are more likely to choose problems involving
difficulty level, students may choose a very easy (ve), easy (ea), addition. These game behaviors can be attributed to the fact that
medium (md), hard (ha), or very hard (vh) game setting. The type the goal of the students is to achieve high scores. It can be also
of problem solved is the type of arithmetic problems (i.e., addition observed from the figure that the distribution probabilities in hare
(add), subtraction (sub), multiplication (mul), division (div)) that hunters are scattered among the 10 states. The finding implies that
the student attempted to solve. Students were informed that prizes hare hunters are solving diverse problems than stag hunters.
109
This study aims to describe the game behavior of
students who utilized a mobile game learning application in
mathematics and predict whether they will exhibit a stag or a hare
hunter game behavior. It is found that there is a very high
probability that students will stay in the game settings they choose
until the end of the game. Students that display a hare hunting
behavior solve a variety of problems in an easy and slow pace
manner. This is different from the students who are stag hunters.
Fig 1. A hidden Markov model for relating the game mode settings Stag hunters are risk-takers – solving more difficult problems also
to the game behavior (stag and hare) with a faster game setting. Thus, we found evidence that stag and
hare hunters of a mobile game learning application have unique
Tables I and II show the three possible cases the students traits that separate the one from the other.
may exhibit during game sessions. The other twenty-seven possible
cases have the same output as the first-three observations (Obs). It The results of this study will be implemented in the
indicates that students are more inclined to be hare hunters – a development of an adaptive version of Ibigkas! Math. The adaptive
"slowly but surely" game behavior. The results shown in Table 1 version will be then tested to determine the changes in the gaming
further confirm that hare hunters solve a variety of problems. The behaviors of the students. Its influence on students’ mathematics
hare hunters will attempt all arithmetic problems with varying learning will also be investigated.
levels of difficulties.
ACKNOWLEDGEMENT
TABLE I. Predicted Game Behaviors Given the Game Settings
N Obs1 Behavior n Obs2 Behavior The researchers are indebted to Dr. Nieva Discipulo,
1 haadd Hare 1 vhadd Stag Ellen Tabayan, UE Elementary and Senior High School Laboratory
2 mdmul Hare 2 vhadd Stag School staff, my research assistants, and to all participants of the
study. This paper is accepted for oral presentation at the
3 mdsub Hare 3 vhadd Stag International Conference on Computers in Education 2020.
4 mdadd Hare 4 vhadd Stag
5 vesub Hare 5 vhadd Stag REFERENCES
6 eadiv Hare 6 vhadd Stag
7 vesub Hare 7 vhadd Stag [1] R. P. Bringula, M. M. T. Rodrigo, J. Ocumpaugh, K.
8 eamul Hare 8 vhadd Stag Porayska-Pomsta, I. Olatunji, and R. Luckin, “Towards the
development of a computer-based game for phonemic
9 vesub Hare 9 vhadd Stag
awareness”, In J. C. Yang et al. (Eds.), Proceedings of the
10 veadd Hare 10 vhadd Stag 26th International Conference on Computers in Education.
Paper presented at the International Conference on Computers
TABLE II. Predicted Game Behaviors Given the Game Settings in Education (pp 657-662). Philippines: Asia-Pacific Society
(continuation) for Computers in Education, 2018.
N Obs2 Behavior [2] R. P. Bringula, M. M. T. Rodrigo, “Can "stag-and-hare hunt”
1 vhadd Stag behavior be modeled using interaction data from a mobile-
2 vhadd Stag supported collaborative learning application?”. In M. Chang
3 vhadd Stag et al. (Eds.), Proceedings of the 27th International Conference
4 veadd Stag on Computers in Education. Paper presented at the
5 vhadd Stag International Conference on Computers in Education (pp 761-
6 eadiv Hare 766). Taiwan: Asia-Pacific Society for Computers in
7 eamul Hare Education, 2019.
8 mdadd Hare [3] R. P. Bringula, M. M. T. Rodrigo, “Modeling stag and hare
9 veadd Hare hunting behaviors” (unpublished manuscript). Ateneo de
10 vesub Hare Manila University, Philippines, 2020.
[4] R. Gillies, “Cooperative learning: Review of research and
practice” Aust. J. of Teach. Edu., vol. 41, no. 3, pp. 39–51,
Meanwhile, a stag hunter tends to solve very hard 2016. doi:10.14221/ajte.2016v41n3.3
addition problems. Perhaps, they believe that they will gain more
points by solving more difficult problems. Consistent with the
definition of stag behavior, the students who chose to set the game
mode in a faster setting are risk-takers. A faster game setting may
incur higher points but the game imposes higher time deductions
for every wrong answer. The HMM model can also predict that
students may also shift from being a stag to a hare hunter (and vice
versa). As previously shown, the shift in a game behavior is very
minimal. Nonetheless, it provides an opportunity for learners to
explore more game settings.
IV. CONCLUSION
110
Marketing and Engagement Tool for Educational
Institutions
John Cristian Barrameda1, a , Eizer Edzon Relayson1,b, Richelle Ann B. Juayong1,c, Jasmine A. Malinao2,d
,
Jaime D.L. Caro1,e, and Jozelle C.Addawe*1,f
1
Service Science and Software Engineering Lab
Department of Computer Science
University of the Philippines Diliman
Quezon City, Metro Manila
2
Headstart Business Solutions, Inc.
Quezon City, Metro Manila
a
jmbarrameda@up.edu.ph, begrelayson@up.edu.ph, crbjuayong@up.edu.ph, djas_malinao@yahoo.com,
e
jdlcaro@up.edu.ph, fjcaddawe@up.edu.ph*
Abstract— Educational institutions find it necessary to make of marketing and communication actions [2]. Making sure that
efforts to promote high student recruitment, minimize student students do not suffer a tendency to get bored, dispassionate,
dropouts and encourage continuous student engagement and disaffected, or otherwise “disengaged” fosters student engagement
alumni engagement. However, in the Philippines, there is currently [3]. Finally, an institution may offer personally relevant benefits
a lack of research that aims to develop a system which focuses to that attract the attention of the alumni, and incentivize them to
address student recruitment, student engagement, and alumni respond in order to encourage alumni engagement [1, 4].
engagement. In this paper, we present a web application system
design that can be used by institutions as a marketing and Some of the factors that influence student recruitment are school
engagement tool designed to promote student recruitment, student visibility, school prestige, and suitable programs for the students
engagement, and alumni engagement of an institution. The features while student retention may be influenced by an institution’s ser-
of the application include an interactive survival guide, a chatbot, a vices and policies, buildings and infrastructures or concerns
Social listener and an account management system. Altogether, the regarding school fees.
system is a timely and cost efficient way to evaluate and maintain
the engagement level of students to an institution even in a remote School visibility covers the marketing strategies of the institution
manner. [5] while school prestige includes faculties’ reputation, government
and private agencies’ recognition, and student’s proximity and
accessibility to the educational institute [6, 7]. Suitable program is
Keywords—education, marketing tool, engagement tool, related to course compatibility, tuition fee affordability, and an
engagement tool, student recruitment, student engagement institution committed to quality [8, 9].
Promoting and fostering each of the key aspects mentioned above In this paper, we propose a design of a web application system for
require some effort to be carried out by an institution. In order to institutions to promote and foster student recruitment, student
attract, engage and convert individuals into new and consistent engagement and alumni engagement of an institution.
enrollments, an institution, university or school, may carry out a set
111
The outline of the paper is as follows: Section II presents our ob-
jectives and considerations in designing the application, in Section
III we present the application design, initial visualizations and
system testing plans. Conclusions are then summarized in Section
IV.
These features were identified by the proponents so that it can Fig. 1. Login Interface
cover important considerations that might influence a student’s
decision making in entering a school and continuously engaging
with the institution.
Open source libraries are used in the creation of the said features.
The map used in creating the Interactive Survival Guide is limited
to the vicinity of the school and the Chatbot is trained using
English phrases from frequently asked questions about the school.
The social listener is limited to Twitter posts.
112
Fig. 3. Interactive Map Menu
C. Chatbot
D. Social Listening
module. It contains a list of marker suggestions made by guests and • Display positive feedback and trending topics about the
other administrators with its description, name, and coordinates. institution from Twitter posts on the Application so that guests to
the application system can view it.
The list is used by administrators as a guide in updating the dataset
that stores the map markers and details of the Interactive Map. • Assess data through the analytics that will be provided by the
application
Administrators of the system can submit a form containing the
coordinates and the label name to suggest a mark for the system Control management of the social listening feature involves a
administrator to add, edit, or delete map markers. Guests/Viewers configure dropdown accessible by the administrator. There will be
may only view the list of requests of the users to add, edit, or delete two options in the dropdown menu:
a mark.
The Guest View page in Fig. 7 displays feedback regarding the
Guests can only view the interactive map and the details associ- institution through an embedded link that will be set up by an
ated to the marked locations and events or view the animation of administrator. Twitter Publish will be used to get the collection of
the institution’s registration process created by the administrator. twitter posts from the institution’s page. The proponents suggest to
use the institution page’s likes collection to be displayed. Em-
bedded twitter links were used in the system due to automation rule
113
Fig 9. System User Flow
(4) User post with the highest retweet and followers, to identify
Fig. 7. Social Listening Guest View Sample Layout people who greatly impact the brand or keyword.
The Admin View page in Fig. 8 displays a collection of tracked The administrator can use the keyword search to indicate which
conversations about a specific phrase, word, or the institution name analytics result will be displayed on the page. Same rules from
in social media posts. The purpose of this page is to provide some Keyword Settings can be applied to Keyword Search.
degree of insight through the analysis of said conversations to be
able to devise a strategy that will better influence targeted After the administrator provides a keyword to the social listening
audiences. This page allows the admin to view Graphs and process, guests are granted access to feedback, mentions, and list
Analytics, access Keyword Settings and use the Keyword Search. of trending topics about the institution in the Guest View page.
The Graph and Analytics allow the administrator to monitor social E. System User Flow
media interactions that include keyword and institution mentions
through summary charts and graphs containing the following Fig. 9 summarizes the user flow of the system and includes the
information: main components used behind the creation of the system features.
114
REFERENCES
115
Semi-immersive Virtual Reality Game for Upper Body
Rehabilitation of Pediatric Cerebral Palsy Patients
Jesrelle Anne S. Abenojar, Kyle Steffi D. Bonifacio, Maria C. Torres
Jaime Caro, Ph.D., Richelle Ann Juayong, Ph.D., Josiah Cyrus Boque*
Service Science and Software Engineering Laboratory
Department of Computer Science
University of the Philippines Diliman
Quezon City, Metro Manila
*jrboque@up.edu.ph
Abstract—Cerebral palsy is a group of permanent and On this research, a semi-immersive VR learning and
nonprogressive disorders that can cause problems in movement, motor rehabili-tation system is created for pediatric patients with cerebral
coordination, and posture. There are traditional rehabilitation methods palsy.This VR shall provide educational activities through
for patients with cerebral palsy, but pediatric patients are often gamification of proposed system’s set of activities for physical and
unmotivated in engaging in these methods since they are arduous. This occupational therapy. Also, it will provide visual feedback to the
research explores the use of virtual reality as a rehabilitation tool for patients’ perfor-mance and improvements. A questionnaire based on
CP patients as it allows simulation of the real world and gamification, ISO standard25022 and 25023 will be used to analyse the usability on
which can give motivation and engagement to the patients. The patients,engagement and enjoyment factor of this VR
researchers of this study proposed a game that uses the Kinect sensor
for rehabilitating the upper body of pediatric cerebral palsy patients.
116
achievements. This gives a sense of achievement to users as VRallows In this paper, level progression is being included as part of
them to perform tasks they cannot do in the real world [5].Reward the games. The results (time completed, failed attempts, range of mo-
systems are used by businesses to create rapport and give exclusivity tion) are computed by the system and are then used to generate the next
to their best customer. level. This is done to avoid stagnation and motivation loss because of
the repetitive game. Nah et al. [13] identifies competition as one of the
Lee and Hammer [10] state that the goal of gamification is principles of gamification. This means that games are aimed to provide
touse the motivation and engagement found in game to apply to real a challenge for users to maintain engagement[13]. A form of
problems, such as health. Additionally, they state that games allow “difficulty progression” is needed to maintain en-gagement and
repeated failure. It invites a safer and more welcoming environment motivation.
for trial and error, which helps in maintaining the engagement of the
user. An issue raised by this research is the use of HMDs in the
VRsimulation. Some patients reported nausea and eye-irritation upon
Several studies explore the application of gamification and using the HMDs within the length of testing [3]. Pereira [16] also
VRto CP rehabilitation. These studies offer different perspectives and discusses nausea due to prolonged exposure to HMDs. This provides
data in developing this research. us the idea of removing the HMD from the proposed game's hardware
as per the comfort of the patients.
One research uses a semi-immersive multi-touch display
screen to simulate the virtual environment in different games [8]. A Subsequently, there is a study that incorporates all of this
motion sensor ball and pen are used for user interaction. The using a Kinect sensor to create a ludic game system for mobility
researchers develop a series of games that are aimed to be used by disorders[17]. The system consists of six games developed using
occupational therapists. One of the games created is called Catching Unity. Their Use of Kinect provides us the idea to use the
Butterflies,which requires the use and coordination of both hands. commercially available device for research. The use of Kinect also
saves the users from wearing body sensors that can cause discomfort
The research states their design process which includes the [5].
study of current related literature, visit of cerebral palsy clinics and
ther-apy sessions, and consultation with medical experts [8]. These are The study also provides the details of the technical part of
considered in the development of the proposed system, especially with developing their system, which includes the data layers. This can be a
the consultation with medical people. guide in designing the architecture of the proposed system.
Another research uses the sit-down setup but instead of Another factor that is being improved upon is the level of
motion sensor balls, the researchers use the Leap Motion device, which progression or the difficulty curve of the activities. The researchers
isa sensor that accepts fine motor input [7]. The system features a develop the games wherein it can be customized by the administer-ing
multiple game setup that allows the patient to pick and drop objects in therapist. Each game has a set of controls to alter the number of
the VR world. The game aims to improve joint flexion in the hands obstacles, distance of each obstacle from each other, and more. This
which is related to fine motor skills. Approach aims to have experience for each patient and therapist more
personalized.
Said games feature colorful environments and animation
mod-els. It also heavily focuses on feedback and positive III. METHODOLOGY
reinforcement as the study aims to maintain user engagement. To
measure user engagement, the researchers use a device called A. Theoretical Framework
MindWave that detects the brain waves of the person wearing the head Figure 1 shows the details of the theoretical framework that dis-cusses
device [7]. the bulk of theories and concepts used to create and support the system.
While they found that the MindWave provides accurate As seen in figure, the patient undergoes pre-assessment to
results in brain wave mapping, they could not make a general claim for deter-mine their GMFCS and MACS levels. These levels will
its efficiency as some of the patients rejected the MindWave [7].The determine if they are included in the scope of the project [9,15]. After
researchers proceed to suggest the use of Kinect sensor for future which, the upper body motor proficiency of the patient will undergo
iterations of the study to increase the scope of movement captured by measure-ment based on the BOT2 guidelines on manual coordination
the system [7]. and body coordination [4]. Exercises for manual coordination consist
of reaching, wrist movement, and hand-eye coordination. While he
The previous studies utilize a sit-down setup in the game exercises for the body coordination part consists of alternating
process.Other researches have a standing setup, most likely to create a movements of the left and right side of the body. In some
more flexible space for the patient to interact in.The next research cases,occupational therapy is used instead of physical therapy. The
decides to integrate wearable haptic device to simulate wrist and finger researchers have listed down some activities that can serve as occu-
movement. These devices are paired with an HMD in a fully pational therapy for the patients [14].
immersive VR setup [3].
117
Fig 1. Theoretical Network
Figure 1 also illustrates the different researches discussed in The therapist can adjust the game parameters based on the
the literature review and the different approaches they used to mo-tor skills of the player. The increase in difficulty will then be based
integrateVR to the rehabilitation process [3,7,8,17]. The proposed on the set parameters. For now, the identified modifiable game
game will incorporate the exercises from the manual coordination parameters are the number of ingredients and the hands to be used(left,
inthe controls and objectives of the game. It will use Kinect as the input right, or both). These are subject to change based on future
and output device that will simulate the VR environment. There- consultations with medical experts.
assessment of the patient will be inputted into the system to keep track
of records. After that, a testing period will occur where testers will The setup of the proposed system is shown in Figure 2. It is
answer two questionnaires; one for usability, which is based on ISO a sit-down setup; the player is seated facing the sensor and themonitor.
standards, and one for engagement, which is based on Fig 2. Proposed Game Set-up
GAMEFULQUEST.
118
Fig. 3. Wireframe for Therapist
Figure 3 shows the wireframe for the therapist. The system Figure 4 shows the wire The game will display a cauldron
will first display an option for viewing the player list, creating a and different ingredients. An Ingredient will glow with a specific color
player,or closing the application. When the first option is chosen, it that indicates which hand will be used. The player is then asked to drag
will lead to the display of all created players. The therapist will then the glowing ingredient towards the cauldron. The dragging of the
choose a player from the list. This will lead to the display of the ingredients simulates the reaching exercises and elbow flexion done in
player’s details with options of editing the details, viewing therecords, traditional therapy settings. The therapist can apply pressure onto the
or starting the game. arm if they want and assist the patient in how long they hold the hand
out. An added difficulty to be inserted if the therapist wishes is the
The edit details option will lead to the scene where the stirring of thepot. The player uses the right and left flicks of the wrist
therapist can change the player’s details. The view records option will to move the ladle. The wrist to be used depends on the color of the
lead to the display of player’s records. The play option will lead to the ladle orto be indicated by the therapist.
scene of setting the game parameters. Then an option for starting the
game is displayed. The game is finished when all the ingredients are
successfully dragged towards the cauldron. It then proceeds to print a
The create player option will lead to a form for creating a compre-hensive report to the therapist, which is also saved as the
newplayer. After it is successfully created, it will display the details of player's record.
the player, an option to go back to the main menu, and a playoption,
which will lead to the set game parameters scene.
C. Hardware
When the therapist chooses to start the game, they will be The proposed system would use Kinect 2 for motion capture.
firstled to an interface where they can set the game parameters that will Itallows full free body motion detection, supports Unity game
suit the motor skills of the player. After setting the game pa-rameters, development, and has available Blender toolkits for motion capture
they can continue to the game. The game will be both displayed on the modeling.
screens of the player and therapist.
119
that the therapist will use will have at least the following specifications
so that the Kinect 2 will properly function:
● 64-bit (x64) processor
● Physical dual-core 3.1 GHz (2 logical cores per physical
sys-tem) or faster processor
● 4 GB of RAM
● Graphics card that supports DirectX 11
● Windows 8 or 8.1, Windows Embedded 8, or Windows 10
Additional hardware necessary for the system:
● USB 3.0 controller dedicated to the Kinect for Windows v2
sensor or the Kinect Adapter for Windows for use with
theKinect for Xbox One sensor
D. Software
The system will be using Blender and Unity for software
development. Blender is a free and open source 3D creation suite used
for modeling, rigging, animation, simulation, rendering, compositing,
motion tracking, video editing and game creation. Unity is a cross-
platform game engine that is compatible with multimedia or game
development software such as Blender, and uses C# and JavaScript.
E. Architectural System
The proposed system consists of three layers.
120
ISO 25022 and ISO 25023 are standards and guidelines to [6] Pietro Cipresso, Irene Alice Chicchi Giglioli, Mariano Alcañiz
be used for measuring usability. Raya, and GiuseppeRiva. 2018. The past, present, and future of
virtual and augmented reality research:a network and cluster
ISO 25022 focuses on the quality of use of the software. analysis of the literature.Frontiers in psychology 9 (2018),2086.
These Include measures in user efficiency and risk management. The [7] Juliana M de Oliveira, Rafael Carneiro G Fernandes, Cristtiano S
focus is on how well the system executes the functionalities, responds Pinto, Plácido RPinheiro, Sidarta Ribeiro, and Victor Hugo C de
to errors, and the overall user satisfaction upon using the system [2]. Albuquerque. 2016. Novel Virtual environment for alternative
treatment of children with cerebral palsy.Computational
ISO 25023 focuses on quality as a product of the software. intelligence and neuroscience 2016 (2016).
These include measures in the maintainability of the system as a [8] Alan Dunne, Son Do-Lenh, Gearóid Ó’Laighin, Chia Shen, and
product.The focus is on how easy the product is to maintain after the Paolo Bonato. 2010.Upper extremity rehabilitation of children
initial release and how intuitive the product is to the intended users [2]. with cerebral palsy using accelerome-ter feedback on a
multitouch display. In2010 Annual International Conference of
These standards will be used to create a questionnaire to be the IEEE Engineering in Medicine and Biology. IEEE, 1751–
dis-tributed to the testers during the project’s testing period. 1754.
[9] Ann-Christin Eliasson, Lena Krumlinde-Sundholm, Birgit
For user engagement, the researchers have decided to create Rösblad, Eva Beckung,Marianne Arner, Ann-Marie Öhrvall, and
a questionnaire that emulates the structure of the GAMEFULQUEST. Peter Rosenbaum. 2006. The ManualAbility Classification
This is a questionnaire that measures the gamefulness of a game system System (MACS) for children with cerebral palsy: scale
based on seven parameters: accomplishment, challenge,guided, development and evidence of validity and
playfulness, competition, immersion, and social [12]. reliability.Developmental medicine and child neurology 48, 7
(2006), 549–554.
IV. CONCLUSION [10] JESSICA Hammer and J Lee. 2011. Gamification in Education:
With the power of VR, therapy for Cerebral Palsy has given a new What, How, WhyBother.Academic Exchange Quarterly 15, 2
dimension for patients to enjoy. Using this new way of rehabilitation, (2011).
patients will be encouraged to do their therapy process, and at the same [11] Angelita Roma P Hebreo, Cynthia D Ang-Muñoz, Joycie Eulah
time, enjoy the use of the technology. With some evidence of E Abiera, Mon-alisa L Dungca, and Betty D Mancao. 2017.
improvement using similar devices for therapy of Cerebral Palsy or Profile of Pediatric Patients withCerebral Palsy at the Department
other mental diseases, it is also expected that this type of VR Therapy of Rehabilitation Medicine, Philippine GeneralHospital. (2017).
for Cerebral Palsy patients will give the same improvement for patients [12] Johan Högberg, Juho Hamari, and Erik Wästlund. 2019. Gameful
mobility especially on their limbs. However, this has not been tested ExperienceQuestionnaire (GAMEFULQUEST): an instrument
to show results of improvement of the patients. for measuring the perceived gamefulness of system use.User
Modeling and User-Adapted Interaction(2019),1–42.
To further improve this research, testing for patients is [13] Fiona Fui-Hoon Nah, Venkata Rajasekhar Telaprolu, Shashank
recommended. Consultation with medical people and therapy centers Rallapalli, andPavani Rallapalli Venkata. 2013. Gamification of
is highly recommended too to prove its conceptual framework with Education Using Computer Games. InHuman Interface and the
medical evidence and theories. Observing a therapy session can also Management of Information. Information andInteraction for
be helpful in gaining knowledge and insights to improve the system Learning, Culture, Collaboration and Business, Sakae
and thus creating more gamified activities that patients can enjoy, and Yamamoto(Ed.). Springer Berlin Heidelberg, Berlin, Heidelberg,
at the same time, improve their situation and way of life. 99–107.
[14] Dilip R Patel. 2005. Therapeutic interventions in cerebral
REFERENCES palsy.The IndianJournal of Pediatrics 72, 11 (2005), 979–983.
[15] Andrea Paulson and Jilda Vargus-Adams. 2017. Overview of
[1] 2019. Role of a Physical Therapist.
four functional classification systems commonly used in cerebral
https://www.apta.org/PTCareers/RoleofaPT/
palsy.Children4, 4 (2017), 30.
[2] Nigel Bevan, Jim Carter, Jonathan Earthy, Thomas Geis, and
[16] E Monge Pereira, F Molina Rueda, IM Alguacil Diego, R Cano
Susan Harker. 2016.New ISO standards for usability, usability
De La Cuerda,Alessandro De Mauro, and JC Miangolarra Page.
reports and usability measures. InInternational Conference on
2014. Use of virtual reality systems as a proprioception method
Human-Computer Interaction. Springer, 268–278.
in cerebral palsy: clinical practice guideline.Neurología (english
[3] Ilaria Bortone, Daniele Leonardis, Massimiliano Solazzi,
edition)29, 9 (2014), 550–559.
Caterina Procopio,Alessandra Crecchi, Luca Bonfiglio, and
[17] A Pérez-Muñoz, P Ingavélez-Guerra, and Y Robles-Bykbaev.
Antonio Frisoli. 2017. Integration of serious games and wearable
2018. New approach of serious games in ludic complements
haptic interfaces for Neuro Rehabilitation of children with
created for rehabilitation therapies in children with disabilities
movement disorders: A feasibility study. In2017 International
using Kinect. In2018 IEEE XXV International Conferenceon
Conference onRehabilitation Robotics (ICORR). IEEE, 1094–
Electronics, Electrical Engineering and Computing
1099.
(INTERCON). IEEE, 1–4.
[4] Robert H Bruininks et al.1978.Bruininks-Oseretsky test of motor
[18] Albert A Rizzo and J Galen Buckwalter. 1997. Virtual reality
proficiency.American Guidance Service Circle Pines, MN.
and cognitive assessment.Virtual Reality in Neuro-Psycho-
[5] Yao-Jen Chang, Wen-Ying Han, and Yu-Chi Tsai. 2013. A
Physiology: Cognitive, Clinical andMethodological Issues in
Kinect-based upper limb rehabilitation system to assist people
Assessment and Rehabilitation 44 (1997), 123.
with cerebral palsy.Research in developmental disabilities 34, 11
[19] Peter Rosenbaum, Nigel Paneth, Alan Leviton, Murray Goldstein,
(2013), 3654–3659.
Martin Bax,Diane Damiano, Bernard Dan, and Bo Jacobsson.
121
2007. A report: the definition and classification of cerebral palsy rehabilitation tool.Journal of neuroengineering and rehabilitation
April 2006.Developmental medicine and child neurology. 1, 1 (2004), 12.
Supplement 109 (2007), 8–14. [22] David M Whittinghill. 2014. Gamification of physical therapy for
[20] Andrew Patrick Hayes Weightman, Nick Preston, Raymond Holt, the treatment of pediatric cerebral palsy: a pilot study examining
Matthew All-sop, Martin Levesley, and Bipinchandra Bhakta. player preferences.age 24(2014), 1.
2010. Engaging children in healthcare technology design:
developing rehabilitation technology for children with cerebral
palsy.Journal of Engineering Design21, 5 (2010), 579–600.
[21] Patrice L Weiss, Debbie Rand, Noomi Katz, and Rachel Kizony.
2004. Videocapture virtual reality as a flexible and effective
122
The Effects of Quality of Service Mechanisms
in a Software-Defined Networking Environment
on HTTP Requests
Josiah Eleazar T. Regencia1, a *, William Emmanuel S. Yu, Ph.D2,b
1
Ateneo de Manila University, Loyola Heights, Quezon City
2
Ateneo de Manila University, Loyola Heights, Quezon City
a
josiah.regencia@obf.ateneo.edu, bwyu@ateneo.edu
Abstract—This study explores how Quality of Service A. Software-Defined Networking and OpenFlow
(QoS) mechanisms using the Software-Defined Networking
distributed approach, Leaf-enforced, affect HTTP requests across Software-Defined Networking (SDN) is a networking
the network on multiple layers. It is important to note that traditional paradigm that introduces a centralized approach to network
approaches commonly enforced QoS in a core switch, Core- management where the control logic of the network is dictated by a
enforced since this is assumed to perform better because it is only controller located inside the control plane which is directly
enforced at a single node. Results in this study have shown that for connected to all packet forwarding networking devices, the data
2 layers and 3 layers of client leaf switches Leaf-enforced QoS plane [4]. The centralized controller is less prone to error when
performed over 30% better than the Core-enforced. Although Core- modifying network policies through high-level languages and
enforced QoS performed better by at least 2.2% in 4 layers of client software components. Furthermore, SDN eases into Network
leaf switches, 5 layers and 6 layers of client leaf switches resulted in Function Virtualization (NFV) which potentially gives major
Leaf-enforced having better performance by at least 1% and 2.8%, advantages in network flexibility, scalability, and also reduction of
respectively. infrastructure costs. Lastly, more fine-grained policies can be
developed and deployed for the purposes of traffic engineering,
Keywords—Software-Defined Networks, Quality of Quality of Service (QoS) provisioning, security and other network
Service, Class-based Queueing, HTTP Traffic management essentials [5]. Multiple studies have been explored on
taking advantage of the distributed nature of SDN to address security
I. INTRODUCTION issues. Peña and Yu developed the use of a distributed firewall for
SDN [6].
The rapid increase in networking applications and OpenFlow is the protocol that enables network operators
devices connected to the internet has opened opportunities to to perform fine-grained implementations of network flows inside the
network researchers to explore the field of Quality of Service (QoS) SDN architecture [7]. Most common SDN controllers use OpenFlow
inside a Software-Defined Networking (SDN) environment, a as the backbone of their framework. Inside every OpenFlow enabled
relatively new technology that decouples the control plane and the switch d is the Flow Table which contains instructions how a switch
data plane. SDN has been regarded as a potential solution to deal is going to process a matched flow [8]. OpenFlow is the protocol
with the challenges traditional networking architecture have with used in this study.
regards to QoS. A previous study by the authors introduced a testing
framework for QoS mechanisms and network topologies inside a B. Quality of Service in SDN
SDN environment [1]. Using the test framework, the authors
explored how layers in the network topology affect QoS Different needs of different types of network flows makes
performance in the network inside an SDN environment [2] and how applying proper QoS mechanisms a challenging task. Prioritiza- tion
additional network traffic flows affect the performance of the SDN and classification are both key factors in enforcing proper QoS
environment enforced with QoS across multiple hops [3]. All studies mechanisms for a certain flow [7]. Civanlar et al. [9] introduced a
involve simulating synthetic HTTP and streaming traffic along with linear programming-based formula to calculate QoS-rich paths
ICMP Ping packets. The study involving additional network traffic where video flows are routed into while best-effort traffic is kept on
flows [3] also included generation of raw IP data traffic with the use shortest paths. OpenQoS [10] uses packet header fields in order to
of the iperf3 tool in order to flood the network with large traffic classify incoming flows as either multimedia flows or data flows
flows. This study extends the previous study [3] by examining how sending multimedia flows to QoS-rich paths while data flows are
QoS mechanisms in the SDN environment affect HTTP requests. In routed using best- effort routing.
this study, the SDN approach is referred to as Leaf-enforced QoS QosFlow [11] manipulates multiple packet schedulers
since QoS is enforced distributedly in client leaf switches. The such as Hierarchical Token Bucket (HTB), Random Early Detection
traditional centralized approach will be referred to as Core- enforced (RED), and Stochastic Fairness Queuing (SFQ) schedulers. Chato
QoS as QoS will only be enforced in the core switch. and Yu [12] [13] also explored the effects Basic Class- based
Queueing (traffic protocol classifications) and Source Class-based
Queueing (source IP address classification). Both mechanisms were
II. RELATED LITERATURE performed under two test cases. The traditional centralized Core-
123
enforced QoS and the distributed SDN approach Leaf-enforced
QoS. CBQ Class Profile CBQ Class Switch QoS
A Previous study by the authors introduced a test Grouping
framework for QoS mechanisms and network topologies in the
context of SDN [1]. Using the introduced test framework, separate Basic at the Leaf Traffic Protocol Client Leaf
studies [2] [3] have been performed in order to explore how layers Switches
in the network topology affect the performance of QoS for both
Leaf-enforced QoS and Core-enforced QoS. This studies involved Basic at the Core Traffic Protocol Core Switch
simulating synthetic HTTP, streaming traffic, and ICMP Ping tests.
Results were then analyzed for IP raw data throughput, HTTP data Source at the Leaf Source IP Address Client Leaf
throughput, streaming bitrate and latency which showed the Switches
advantage of SDN over the traditional centralized QoS enforcement.
Source at the Core Source IP Address Core Switch
Despite having more nodes enforced with QoS in the network,
results have shown that Leaf-enforced QoS generally performed
Destination at the Destination IP Client Leaf
better than Core-enforced QoS. It is important to note that the Core-
Leaf Address Switches
enforced has always been assumed to perform better since it has only
one node enforced with QoS, hence, commonly used. However,
Destination at the Destination IP Core Switch
when more network traffic was generated with the use of iperf3,
Core Address
results showed that all cases of SDN approach QoS were
significantly better than the centralized approach for both raw IP
Source-Destination Source-Destination Client Leaf
data throughput and latency [3].
at the Leaf IP Address Switches
124
1) a diversity of traffic types across the network,
2) and in order to have large traffic flows flowing across the
network.
The large traffic flows is necessary as traffic flowing across today’s
networks increases at a rapid pace.
125
Fig. 6. 5 Layers Results
The case for 4 layers of client leaf switches, Figure 6, is TABLE II. 4 LAYERS COMPLETED REQUESTS
the only case where Core-enforced QoS generally performed better P-VALUES.
than Leaf-enforced QoS. Only Basic CBQ Leaf-enforced QoS
case Basic Basic Src Src Dst Dst Src- Src-
resulted in having more Completed Requests than its Core-enforced
at at at at at at Dst at Dst at
QoS counterpart, approximately 18% more Completed Requests.
Leaf Core Leaf Core Leaf Core Leaf Core
The rest showed that the Core-enforced QoS mechanism has more
Completed Requests. However, the p- values for this case in Table I Basic at 1.0 0.350 0.461 0.975 0.531 0.967 0.942 0.967
shows that no p-value is less than α=0.05. Hence, the difference is Leaf 1 8 8 9 3 5
insignificant and the Leaf- enforced QoS still perform as good as
their Core-enforced QoS counterparts. In the same way, where Core- Basic at 1.0 0.834 0.328 0.762 0.357 0.378 0.332
enforced QoS all have less Failed Requests than Leaf-enforced QoS, Core 3 4 7 2 3
the p-values in Table II also show that the differences are sta-
Src at 1.0 0.437 0.921 0.474 0.498 0.439
tistically insignificant. Hence, despite having less Completed
Leaf 3 5 7 1 7
Requests and more Failed Requests, Leaf-enforced QoS still
perform as good as Core-enforced QoS for HTTP requests in the Src at 1.0 0.506 0.942 0.917 0.991
case of 4 layers of client leaf switches. Core 8 2
For 5 layers of client leaf switches, Leaf-enforced QoS
mechanisms once again show to have more Completed Re- quests Dst at 1.0 0.547 0.571 0.506
than the Core-enforced QoS mechanisms counterparts. Although Leaf 1 9
Leaf-enforced Basic CBQ QoS and Leaf-enforced Destination CBQ Dst at 1.0 0.973 0.934
QoS only have 0.5% and 1.7%, respec- tively, more Completed Core 5 9
Requests than their Core-enforced QoS counterparts, Leaf-enforced
Source CBQ QoS and Leaf- enforced Source-Destination CBQ QoS. Src-Dst 1.0 0.909
For Failed Requests, once again Core-enforced QoS have less Failed at Leaf 7
Requests than Leaf-enforced QoS. But the p-value in this case was
0.9735 which is less than α=0.05, hence statistically insignificant. Src-Dst 1.0
For 6 Layers of client leaf switches, despite Core-enforced at Core
Source CBQ and Core-enforced Source-Destination CBQ hav- ing
more Completed Requests than their Leaf-enforced QoS
counterparts, this is shown to be statistically insignificant in Table
V as no p-value is less than α=0.05. Moreover, Leaf- enforced Basic
CBQ QoS and Leaf-enforced Destination CBQ QoS have 16% and
4%, respectively more Completed Requests than their Core-
enforced QoS counterparts. With regards to the Failed Requests, the
Core-enforced Destination CBQ QoS having less Failed Requests
compared to the Leaf-enforced Destination CBQ QoS is shown to
be statistically significant with p-value 0.0309. This is shown in
Table VI.
126
TABLE III. 2 LAYERS FAILED REQUESTS case Basic Basic Src Src Dst Dst Src- Src-
P-VALUES. at at at at at at Dst at Dst at
case Basic Basic Src Src Dst Dst Src- Src- Leaf Core Leaf Core Leaf Core Leaf Core
at at at at at at Dst at Dst at Basic at 1.0 0.409 0.340 0.355 0.309 0.200 0.541 0.604
Leaf Core Leaf Core Leaf Core Leaf Core Leaf 1 3 9 6 3 4
Basic at 1.0 0.846 0.756 0.531 0.579 0.783 0.585 0.521 Basic at 1.0 0.885 0.922 0.814 0.626 0.871 0.759
Leaf 7 7 7 2 1 7 Core 7 7 4 8 9 6
Basic at 1.0 0.611 0.659 0.449 0.627 0.454 0.640 Src at 1.0 0.960 0.924 0.735 0.770 0.657
Core 1 9 8 6 2 Leaf 8 2 9 9 8
Src at 1.0 0.344 0.805 0.960 0.815 0.345 Src at 1.0 0.885 0.693 0.801 0.686
Leaf 9 5 2 4 1 Core 2 4 9 9
Src at 1.0 0.233 0.341 0.233 0.951 Dst at 1.0 0.819 0.710 0.601
Core 5 6 2 Leaf 8 6
Dst at 1.0 0.759 0.987 0.238 Dst at 1.0 0.546 0.436
Leaf 1 9 Core 4 5
Dst at 1.0 0.768 0.343 Src-Dst 1.0 0.901
Core 7 3 at Leaf 8
Src-Dst 1.0 0.238 Src-Dst 1.0
at Leaf 5 at Core
Src-Dst 1.0
at Core
TABLE VI. 6 LAYERS FAILED REQUESTS
P-VALUES.
Fig. 7. 6 Layers Results
case Basic Basic Src Src Dst Dst Src- Src-
at at at at at at Dst at Dst at
TABLE IV. 4 LAYERS FAILED REQUESTS Leaf Core Leaf Core Leaf Core Leaf Core
P-VALUES.
Basic at 1.0 0.196 0.576 0.447 0.032 0.955 0.011 0.238
case Basic Basic Src Src Dst Dst Src- Src- Leaf 5 2 2 8 3 8 8
at at at at at at Dst at Dst at
Leaf Core Leaf Core Leaf Core Leaf Core Basic at 1.0 0.054 0.591 0.349 0.197 0.182 1.0
Core 6 7 4 1 4
Basic at 1.0 0.866 0.799 0.267 0.209 0.106 0.611 0.190
Leaf 7 4 3 2 1 5 Src at 1.0 0.172 0.006 0.517 0.001 0.083
Leaf 3 1 3 7 2
Basic at 1.0 0.936 0.353 0.279 0.149 0.738 0.253
Core 7 3 4 5 1 Src at 1.0 0.150 0.461 0.067 0.624
Core 7 4 2 2
Src at 1.0 0.375 0.295 0.154 0.790 0.266
Leaf 5 5 9 Dst at 1.0 0.030 0.701 0.386
Leaf 9 3 3
Src at 1.0 0.833 0.548 0.542 0.752
Core 8 2 2 Dst at 1.0 0.010 0.242
Core 6 8
Dst at 1.0 0.710 0.433 0.911
Leaf 5 9 1 Src-Dst 1.0 0.216
at Leaf 3
Dst at 1.0 0.246 0.809
Core 8 3 Src-Dst 1.0
at Core
Src-Dst 1.0 0.390
at Leaf 9
VI. CONCLUSION
Src-Dst 1.0
at Core
This study was performed in order to analyze the effects
of Quality of Service mechanisms in a Software-Defined
TABLE V. 6 LAYERS COMPLETED REQUESTS Networking architecture with regards to HTTP requests.
P-VALUES. Simulations in this study also included VLC streaming, ICMP Ping
127
latency tests, and also iperf3 raw IP data traffic generation. These [6] J. G. V. Pena and W. E. Yu, “Development of a distributed
tests were also done in multiple layers of client leaf switches where firewall using software defined networking technology,” in
QoS is distributedly enforced. The purpose was to replicate a real- 2014 4th IEEE In- ternational Conference on Information
world networking scenario where network traffic is large and travel Science and Technology. IEEE, 2014, pp. 449–452.
across multiple hops in order to reach the target destination. Results [7] M. Karakus and A. Durresi, “Quality of service (qos) in
in this study have shown that in general, more HTTP requests were software defined networking (sdn): A survey,” Journal of
completed with the Leaf-enforced QoS. Despite that, results have Network and Computer Applications, vol. 80, pp. 200–218,
also shown that there are more Failed Requests with Leaf-enforced 2017.
QoS for 2 layer, 4 layers and 6 layers of client leaf switches cases. [8] N. McKeown, T. Anderson, H. Balakrishnan, G. Parulkar, L.
However, based on a student’s two-sample t-tests were performed in Peterson, J. Rexford, S. Shenker, and J. Turner, “Openflow:
order to test the statistical significance of the advantage of Core- enabling innovation in campus networks,” ACM SIGCOMM
enforced, results showed that in general the advantage was Computer Communication Review, vol. 38, no. 2, pp. 69–74,
statistically insignificant except for only two Leaf-enforced QoS 2008.
mechanisms in 6 layers of client leaf switches case where [9] S. Civanlar, M. Parlakisik, A. M. Tekalp, B. Gorkemli, B.
Destination CBQ at the Core performed better than Destination CBQ Kaytaz, and E. Onem, “A qos-enabled openflow environment
at the Leaf and Source- Destination CBQ at the Leaf. Furthermore, for scalable video streaming,” in 2010 IEEE Globecom
it was observed in the results that as layers in the network topology Workshops. IEEE, 2010, pp. 351– 356.
increased, the Completed Requests for Leaf-enforced QoS relatively [10] H. E. Egilmez, S. T. Dane, K. T. Bagci, and A. M. Tekalp,
remained the same but the Completed Requests for Core- enforced “Openqos: An openflow controller design for multimedia
QoS steadily increased making the gap from Leaf- enforced QoS delivery with end-to-end quality of service over software-
smaller. Despite that, Leaf-enforced QoS still resulted with more defined networks,” in Proceedings of the 2012 Asia Pacific
Completed Requests than Core-enforced QoS. This goes to show the signal and information processing association annual summit
advantage of the distributed nature of Software-Defined Networking and conference. IEEE, 2012, pp. 1–8.
despite having more nodes in the network enforced with QoS. [11] A.Ishimori,F.Farias,E.Cerqueira,andA.Abele ́m,“Controlofmu
ltiple packet schedulers for improving qos on openflow/sdn
networking,” in
2013SecondEuropeanWorkshoponSoftwareDefinedNetworks.
ACKNOWLEDGEMENT IEEE, 2013, pp. 81–86.
[12] O. Chato and W. E. S. Yu, “An exploration of various qos
The completion of this study has been made possible with mechanisms in an openflow and sdn environment,” in Accepted
the support and guidance of the following people, to whom we for Presentation in The International Conference on Systems
extend our deep gratitude and appreciation: The faculty and staff of and Informatics (ICSAI-2016), 2016.
Ateneo De Manila University Department of Information Systems [13] O. Chato and W. E. S. Yu,, “An exploration of various quality
and Computer Science for support and guidance and our families, of service mechanisms in an openflow and software defined
for all their support, understanding and inspiration. networking environment in terms of latency performance,” in
Finally, we thank the Lord God Almighty without whom 2016 International Conference on Information Science and
none of these would have been possible. Security (ICISS). IEEE, 2016,
REFERENCES
128
Philippine License Plate Localization Using Genetic
Algorithm and Feature Extraction
Patrick Matthew J. Chan, John Anthony Jose, Rhen Anjerome Bedruz, Elmer P. Dadios
Electronics and Communications Enginnering Department
De La Salle University, Manila
patrick_chan@dlsu.edu.ph, rhen.bedruz@dlsu.edu.ph, john.anthony.jose@dlsu.edu.ph, elmer.dadios@dlsu.edu.ph
Abstract— This study aims to focus on the license plate Now, in order to be able to achieve this feature, two main
localization portion of an automated license plate recognition systems are needed as its components, namely: a plate
scheme. In order to achieve this, the study makes use of a genetic localization system, and an optical character recognition (OCR)
algorithm in order to perform a more systematic search over the system [7]. The plate localization system’s purpose is to search
input image; as well as Histogram of Oriented Gradients feature for and isolate potential license plates found within a frame,
extraction technique in order to form the fitness function of the whose output will then be fed to an OCR, such as in [8], for a
genetic algorithm. When the algorithm was tested, it was able to more robust and accurate character recognition within individual
successfully localize the license plates within each of the multiple plate images.
input images. In fact, even though the study was focused on 2014
series license plates, the program results also show that it was also In this paper, the study will be focusing on License Plate
able to roughly localize a license plate from another series as well. Localization, using Genetic Algorithm (GA), as well as feature
Keywords— Genetic Algorithm, Localization, HOG, Feature extraction using Histogram of Oriented Gradients (HOG). In this
Extraction paper, Section II would give a brief discussion of genetic
algorithm, on which the system algorithm in this study is based
I. INTRODUCTION
upon. Meanwhile, Section III would further explain the
approach that was used in the study to successfully perform the
For the Philippines, traffic congestion has been a rampant plate localization system, Section IV would be discussing the
problem, and one of the solutions that the government has results of the study, while Section V would provide overall
implemented to alleviate this, is a contactless apprehension conclusions and recommendations after performing the study.
system that aims to reduce road blockages caused by traffic
violators and poor road discipline. Though numerous cameras
are already in place in order to detect traffic violators from a II. GENETIC ALGORITHM
control center, the recognition of license plates, which is a very
important aspect of contactless apprehension, is still done by Generally, genetic algorithms are typically used to solve
human eyesight, which proves to be inefficient. Currently, there optimization and/or search problems, by simply taking
are various available studies in automating the detection and inspiration on biological events that lead to the evolution of a
apprehension of traffic violators, such as [1] [2] [3], but in species. In the scientific theory of evolution by natural selection,
aiming towards a fully automated contactless apprehension a generation can be seen to consist of stronger individuals and
system, it is important to know the identity of the vehicle itself weaker individuals, that each compete for the same limited
and its owner as well. Thus, the first tool needed to track down resources. As a result, stronger individuals are more likely to
and log violators, is to automate the process of reading license survive, compared to the weaker individuals. As enough time
plates from the incoming video feeds. passes by, the surviving individuals start to reproduce, and the
genetic traits that comprise the surviving individuals are
In fact, aside from smart traffic apprehension systems, randomly combined and passed on to the children. This makes
automated reading of license plates, can even be combined with the child generation to be mostly comprised of individuals that
other algorithms, in order to automated vehicle management mostly have one of the stronger parents’ genes. Then, as the
systems as well. An example would be having a fully automated, children form the next generation, generations continue to
and virtually unmonitored parking system management, such as generally get stronger and stronger as time passes by. This
in [4] [5]. Another possible combination, would be in a fully concept of each generation getting stronger than the previous
automated and tagless Philippine toll gate system, when one is the general idea behind the concept of evolution.
combined with vehicle classification such as [6]
129
In applying these concepts to computer algorithms, a height of 1080 pixels, then the chromosomes can have values
problem is instead solved using a first generation of randomly ranging from (1,1), up until x=1920-91=1829, and y=1080-
generated candidate solutions. Each of this candidate solution, 33=1047; or simply written as (1829, 1047). then starts to
is called a chromosome, which is a data structure that contains perform the genetic algorithm.
the necessary information to be plugged in, like variable values,
to the specified problem. Similarly, in this process of plugging
in each chromosome into the problem, some solutions would
evidently give better results, thus making it a “stronger”
chromosome. Now, this method of knowing how strong a
chromosome is, is called a fitness function, which takes a
chromosome as an input, and takes a numerical value as an
output, indicating the strength of that input chromosome.
Meanwhile, in terms of having each chromosome compete
against one another, this process is called selection, in genetic
algorithm terminology. After this process, similarly, the
chromosomes need to “reproduce”, in order to create the next
generation. At this point, the process of “reproduction” is mainly
designed by the programmer, on how two (or more) parents
produce a new “child” chromosome. In genetic algorithm
terminology, this process of “reproduction” is called crossover.
Lastly, it is important to note that some of the strong traits in
currently evolved creatures in biology, resulted in a mutation, or
a mistake in the copying of genes, leading to new behavior.
Similarly, in genetic algorithm, mutation can also be defined by
programmer, usually by manually editing the values inside a
chromosome, along with a mutation probability that defines how
often this kind of chromosome alteration is likely to happen.
III. METHODOLOGY Fig. 1. System Algorithm Flowchart showing Genetic Algorithm
For the plate localization system, a Genetic Algorithm was
utilized to perform the localization process. With this, the idea
is that various chromosomes, each representing different B. Fitness Function
cropped parts of the input image, will be competing against one
Another important element in the design of a Genetic
another on how much each looks like a license plate; and, as the
Algorithm, is its fitness function. For this study, in relation to
winners “reproduce” and make the next set of chromosomes, the
license plates, although there is no single exact image depiction
whole population will keep evolving, until it eventually
of a license plate, there is a certain set of features that can instead
produces one, that points to the actual position of the plate
be used to describe one. Typically, one can say that a license
number within the picture.
plate is a rectangular-shaped and usually plainly colored flat
In order to further describe how the program algorithm object, containing huge numbers and/or letters in the middle area
works, a high-level flowchart of the general processes involved in easily readable font. Thus, in order to take advantage of plate
are shown in Figure 1. In the first step, the system basically numbers having a common set of features, this study used HOG
reads the image input, and loads it into memory for processing. feature descriptors, as a way to check whether parts of the input
For simplification, the plate size in the image was set to a image resembles a license plate, or not, based on the features of
constant value of 91-by-33. Now, after the image is loaded, the a known license plate. With this in mind, the general idea behind
machine then starts to perform the genetic algorithm. the fitness function used in this study is the difference in HOG
features between a basis image, which is a sample image
A. Chromosome containing only a plate number by itself, and a specific part of
For the chromosomes used in the system, each of them the input image, as specified by a chromosome.
simply contained two decimal values, which denoted x and y
coordinates. These coordinates supposedly point to where the Now, in order to be able to obtain a numerical value for the
top-left corner pixel of the plate number is, in the picture, fitness function, the process begins with first extracting the
according to that chromosome. For this coordinate system, the HOG features of both images, each of which result in a huge
top leftmost pixel is described as (1,1), with the x-coordinate array of normalized numbers, representing the HOG description
increasing in value as one goes rightward, and with the y- of that image. Then, in order to obtain how far or how close the
coordinate increasing in value as one goes downward instead. features of the two images are from one another, a simple
Thus, coordinate values can range from one, to the lower computation using L2 distance, or also known as Euclidean
rightmost pixel within an image, wherein one can still draw a distance, is used, taking the square root of the sum of squared
full-sized license plate rectangle from, thus covering every differences, between corresponding pairs of numbers, between
possible spot within the image, to be a license plate. For the two sets of extracted features. This effectively condenses the
example, if the input image size has a width of 1920, and a two huge arrays of numbers representing two pictures, to a
130
single value, representing the error between thee features of the where 𝑃𝑗 is the probability of chromosome j to be removed,
two images; thus, successfully giving us a numerical value for 𝐹𝑖𝑡𝑛𝑒𝑠𝑠𝑗 is the Fitness value of chromosome j, and 𝑃𝑜𝑝𝑆𝑧 is the
our fitness function. Hence, the fitness function for this genetic population size of the generation.
algorithm is described as:
This easily makes sense, since, as mentioned before, a higher
fitness value represents a more unfit chromosome; thus, making
𝑁
2 it more likely to be removed from the current generation. This
𝐹𝑖𝑡𝑛𝑒𝑠𝑠 = √∑ (𝑓𝑏𝑖 − 𝑓𝑐𝑖 ) ()
𝑖=1 part of the algorithm continues to remove chromosomes from
the population, until only around half of the original population
has remained from the generation.
where N is the total number of elements of the HOG array,
𝑓𝑏𝑖 is the value of HOG feature number i of the basis image, and D. Crossover and Mutation
𝑓𝑐𝑖 is the value of HOG feature number i of the image described After selection has finished, the remaining population,
by the chromosome. including the fittest chromosome, is left to “reproduce”, in order
to create the next generation. For this system, the two parents of
Note that both the basis image and the image described by
each child chromosome were randomly chosen, although a rule
the chromosome should be of the same size, in order to produce
was placed that prohibited a single chromosome to become both
the same number of elements in the HOG feature number array,
parents of a child. Now, since each of the chromosomes contain
in order for this fitness function to successfully give an
x and y coordinate information, the crossover method used in
appropriate value. In addition, it should also be noted that in the
this study involved computing a weighted center point between
system, a lower fitness value indicates a fitter chromosome,
the two parents, based on their individual fitness values,
since the fitness function measures the amount of error between
resulting with the child chromosome’s x and y coordinates being
the basis image and the image pointed at by the chromosome.
slightly closer to the fitter of the two parents in the image.
Now, going back into the flowchart, after the image is
After this crossover process, the mutation possibility of each
loaded, the initial population is generated, giving us a number of
child immediately follows. For this study, the researcher wanted
randomly generated ordered pairs of numbers, depending on the
an exponential formula that greatly increased the chances of
initial population size. After this, each chromosome of the
mutation as the parent chromosomes’ coordinates get very close
current generation, which in this case is the initial randomly
to one another in the image. This was done, in order to avoid the
generated one, is evaluated for its corresponding fitness value,
genetic algorithm from converging to a single value as
mainly based on the portion of the input image each of them
generations pass by, leaving room for enough diversity of
represent. After fitness evaluation, the program then checks for
chromosomes in the next generation as the remnants of the
the stopping condition. In this study, the stopping condition used
parent generation becomes too close to one another in the image.
was when either the fittest chromosome gets a fitness value of
The mutation process is done as follows: first, a “chance” value
zero, or when a set amount of child generations have been
is computed based on the distance between the two parents; then,
generated.
that “chance” value is applied twice for the child chromosome
C. Selection on whether it is to be mutated or not, with one involved with
Whenever the stopping condition is not met, the selection changing the x-coordinate to a random value, and the other
process is done over the current generation, in order to prepare involved with changing the y-coordinate to a random value.
it for making the next generation. In this Plate Localization Thus, for example, if the “chance” value is computed to be 20%,
Genetic Algorithm, the selection process is done through fitness then the actual probability of the child to have a single mutation,
proportionate selection, which is also known as roulette wheel is [ 0.2*(1-0.2)+(1-0.2)*0.2]=0.32, or 32%. On the other hand,
selection, wherein the probability of a chromosome to be the actual probability of the child to have a double mutation,
removed, or “killed”, from a generation is directly proportional (which means that both x and y coordinate values are fully
to how unfit the chromosome is, when compared to all other changed,) is 0.2*0.2=0.04, or 4%. Moreover, the formula used
chromosomes in the same generation. However, a slight to obtain the “chance” value is written as:
modification was made in implementing this genetic operator in
2 −(𝑦 −𝑦 )2 )
this study, such that the fittest chromosome is guaranteed to be chance = (1.5)(10−√(𝑥2 −𝑥1 ) 2 1 % ()
retained, by excluding it in the selection process. This is done,
in order to eliminate the possibility of the algorithm to “forget” where(𝑥1 , 𝑦1 ) and (𝑥2 , 𝑦2 )are the x and y chromosomes of
the best solution as the generations pass by. Thus, in terms of the the two parents.
fitness values explained earlier in this study, the probability of
each non-best-fit chromosome to be removed from the current This formula was chosen, so that the mutation chance would
population is expressed as: only be low, at values less then 1.5%, up until the two parents
become closer than 10 pixels apart, then quickly rises in value.
𝐹𝑖𝑡𝑛𝑒𝑠𝑠𝑗 It should be noted that a maximum chance value of 25% was
𝑃𝑗 = (𝑃𝑜𝑝𝑆𝑧−1) () implemented in the program for this study as well.
∑𝑘=1 𝐹𝑖𝑡𝑛𝑒𝑠𝑠𝑘
Furthermore, another feature that is placed during the
crossover and mutation process, was that the best performer of
the parent generation is retained into the child generation. This
means that a slot in the child population is actually reserved for
131
the clone of the previous best performer. This is, once again,
done in order to guarantee that the best solution over the whole
run of the algorithm, gets passed on until the end, wherein the
stopping condition is met, and the best performer is only taken
from the latest generation as the GA plate localization program
output..
IV. DATA AND RESULTS
For this study, the License Plate Localization System using
Genetic Algorithm was implemented using MATLAB. In
testing the program algorithm with different image inputs, for
the sake of this study, the population size per generation was set
to a fixed value of 250 chromosomes for all runs, while the
Fig. 2. Program Output
number of child generations was set to 1000 generations. These
numbers are used, since it was experimentally noticed that it is
slightly more efficient to allocate a bit more computation time
to producing more generations, rather than when compared to
focusing on the population size over number of generations.
Having a constant setting was done, in order to focus more on
observing examples on how close the program can get to
Fig. 3. Cropped Plate Number from Original Picture
localizing the plate number within the picture, even though the
basis plate image stays the same.
The result of the first run is shown in Figure 2. In here, it can
be confirmed by manual inspection that the system has
successfully localized the plate number within the image. Of
course, in terms of program data, this output is simply the fittest
chromosome of the latest generation, which is simply
represented as the coordinates (1011,474) of a 1920 by 1080
image in this case and has a computed fitness value of around
0.0228. Thus, the program output visualization in Figure 2
simply drew a bounding rectangle over the original input image,
by simply knowing the top left pixel of the output chromosome,
as well as the plate size for the rectangle dimensions. Similarly,
when instructed to only extract the plate number from the picture
from the chromosome, MATLAB also displayed Figure 3 as an
output.
In Figure 4, it shows that the most fit chromosome
practically stayed as is at around generation 81 of the whole
1000-generation run. It can also be observed here that even
though the best fit of each generation seems to be stable, either Fig. 4. Best One’s Fitness vs. Generation (Full, and Zoomed)
steadily being the same value, or going down (or getting fitter);
while on the other hand, we get the second most fit of each In the visualizations on Figure 5 and Figure 6, one can see
generation to somewhat be a fluctuating value as the generations observe as the search process is done over the whole image.
pass by. This property is actually a direct result of the Obviously, unless by lucky chance, the randomly generated
modifications done in the selection, crossover and mutation initial generation would not immediately be able to find the plate
phases as discussed earlier that guarantees the preservation of number immediately within the image. This is only achieved as
the best chromosome overall so far, over future generations. fitter chromosomes are found as generations pass by. So,
Without the application of this rule, one can basically see its following the new best chromosomes as displayed in Figure 5,
effect on the second best of each generation, on which the value one can observe the corresponding “drops” in Figure 4, in terms
can greatly differ over various generations, depending on the of how it gradually moves to spots in the image that resemble a
results of the generation of each of the next generation with the license plate more, all thanks to the HOG feature descriptor as a
crossover and mutation processes. metric indicating this closeness. On the other hand, the
fluctuating values of the second best in Figure 4 is also reflected
in Figure 6, where various parts of the image became second
fittest of its generation, though with most of the ones at the end
being around the plate number as well, as the algorithm just
cannot find a spot within the image that is fitter than the ones
closer to the actual license plate, which is as intended of the
system program.
132
leading to around 80 to 110 of the whole 250-chromosome
population to actually have single mutations, and around 10 to
20 of the whole 250 to actually have double mutations;
therefore, proving the diversity of the later generations. As what
was said earlier, the added diversity by thee mutated children
just happened to perform too badly, that they are practically cut
off from the recombination process by their selection
probability.
133
program algorithm eventually preferred the plate number on the
right over the one on the left by the end of the whole run, due to
the one on the right being of the same series, and thus bearing
more similar characteristics and features.
For this program run described by the program output
displayed in Figure 13, one can see from the program graphs in
Figure 15 that the program did not get to its final chromosome
location as fast as it did during the first run shown in this study
in Figure 4; with this run being only able to obtain a final fitness
value of 8.4503 at the 503rd generation. Even so, the average
fitness graph on the lower left shows that the populations past
this point still ended up being very diverse, even after the
selection process has been done. So, as a result, the “chance”
values ended up only within the bounds of around 0.01, or 1%,
and 0.12, or 12%, by the end of the run; thus, resulting in
mutated children to be only limited to around 0 to 45 of the
whole 250-chromosome population, over the whole program
run.
134
chromosome settled at the 23rd generation, and the final fitness V. CONCLUSION
value was 7.7916. This paper shows that proposed system algorithm was able
to successfully localize the license plate within different input
pictures, even though the basis image stayed the same
throughout the study, which was a 2014 series Philippine license
plate. In fact, it was even able to roughly localize some license
plates from different Philippine license plate series as well,
despite the series limitation in the basis image.
Fig. 13. Program Output (Two different plates) Fig. 16. Detection of Plate from a Different Series
135
local minima as well, and be able to determine whether another convolutional neural network,” 2018 IEEE International Conference on
license plate exists within the picture as well. Applied System Invention (ICASI), 2018.
[3] R. K. C. Billones, A. A. Bandala, E. Sybingco, L. A. G. Lim, and E. P.
Furthermore, the robustness of the algorithm may be further Dadios, “Intelligent system architecture for a vision-based contactless
improved by using a basis image which consist of the average of apprehension of traffic violations,” 2016 IEEE Region 10 Conference
multiple HOG descriptor feature from the same license plate (TENCON), 2016. (null) G. Eason, B. Noble, and I. N. Sneddon, “On
certain integrals of Lipschitz-Hankel type involving products of Bessel
series, for a more exclusive purpose, or for all license plate functions,” Phil. Trans. Roy. Soc. London, vol. A247, pp. 529–551, April
series, to act as more of a one-size-fits-all solution. 1955. (references)
[4] J. Lin, S.-Y. Chen, C.-Y. Chang, and G. Chen, “SPA: Smart Parking
ACKNOWLEDGMENT Algorithm Based on Driver Behavior and Parking Traffic Predictions,”
The authors would like to recognize the Engineering for IEEE Access, vol. 7, pp. 34275–34288, 2019.
Research and Development for Technology (ERDT) Program of [5] Y. He and Y. Jiang, “The Design and Implementation of Residential
Parking Spaces Management and Information Issuing System,” 2013
Department of Science and Technology (DOST) for funding this Fourth International Conference on Digital Manufacturing &
study. Automation, 2013.
[6] R. N. Celso, Z. B. Ting, D. J. R. D. Carmen, and R. D. Cajote, “Two-Step
REFERENCES Vehicle Classification System for Traffic Monitoring in the Philippines,”
[1] R. A. Bedruz, A. C. P. Uy, A. R. Quiros, R. K. Billones, E. Sybingco, A. TENCON 2018 - 2018 IEEE Region 10 Conference, 2018.
Bandala, and E. P. Dadios, “A Robotic Model Approach of an Automated [7] P. Szilveszter and G. Csongor, "Parking surveillance and number plate
Traffic Violation Detection System with Apprehension,” 2018 IEEE 10th recognition application," in Intelligent Systems and Informatics (SISY),
International Conference on Humanoid, Nanotechnology, Information 2010
Technology,Communication and Control, Environment and Management [8] A. R. F. Quiros, R. A. Bedruz, A. C. Uy, A. Abad, A. Bandala, E. P.
(HNICEM), 2018. Dadios, A. Fernando, and D. L. Salle, “A kNN-based approach for the
[2] E. A. Roxas, J. N. Acilo, R. R. P. Vicerra, E. P. Dadios, and A. A. machine vision of character recognition of license plate numbers,”
Bandala, “Vision-based traffic sign compliance evaluation using TENCON 2017 - 2017 IEEE Region 10 Conference, 2017.
136
Evolution of Social Networks in a Social Learning
Management System
Orven E. Llantos1,2,a *, Maria Regina Justina E. Estuar1,b
1
Department of Information Systems and Computer Science, Ateneo de Manila University, Quezon City
2
Department of Computer Science, Mindanao State University-Iligan Institute of Technology, Iligan City
a
orven.llantos@g.msuiite.edu.ph, brestuar@ateneo.ph
137
The agentxagent and agentxtask matrices are converted as V. FUTURE STUDIES
entries to CASOS-ORA XML definition file. Each entry for
initiator, receiver, and task is encoded as nodes for the network The promotion of the integration of online learning in
and the edges as part of the agentxagent and agentxtask matrices. pedagogy to beneficiary schools in the DepEd is timely,
especially during these challenging times of the coronavirus
D. Pattern Interaction Analysis disease 2019 (COVID-19) pandemic. It would be interesting how
the network evolves during these challenging times.
Pattern Interaction Analysis (PIA) is a category of SNA
that studies the interaction patterns derived from the sLMS
ACKNOWLEDGEMENT
networks. There are two main approaches to study PIA that are
inherent in SNA: mathematical analysis and visualization. The
mathematical analysis includes computations for centrality The authors would like to acknowledge DOST-ERDT
measures, density, cliques, and clusters. Mathematical analysis is for the financial support provided for this study, DepEd Iligan,
useful in discovering the structural properties of networks and and Tambo Central School for participating in this study.
communities.
REFERENCES
The SNA tools provide all the information needed to [1] A. Amory, W. Gachie, P. Clarke, and S. Yates, “Needs analysis,
conduct PIA both in mathematical analysis and visualization. PIA development and hci evaluation of a new learning management
allows for the discovery of preferences for communication system,” in Proceedings of EdMedia + Innovate Learning 2004.
Association for the Advancement of Computing in Education
channels, participation levels, collaboration patterns, and (AACE), 2004, pp. 373–380. [Online]. Available:
members contributions. https://www.learntechlib.org/p/12960
[2] M. Raspopovic, S. Cvetanovic, I. Medan, and D. Ljubojevic, “The
III. RESULTS AND DISCUSSION effects of integrating social learning environment with online
learning,” International Review of Research in Open and
To look at how the network evolves, Table I details in Distributed Learning, vol. 18, no. 1, pp. 141–160, 02 2017.
each social network, the density, links, and node count. [3] H. M. Mora, M. T. S. Pont, G. D. M. Casado, and V. G. Iglesias,
“Management of social networks in the educational process,”
Computers in Human Behavior, vol. 51, pp. 890 – 895, 2015.
TABLE I: Network Evolution
[Online]. Available:
http://www.sciencedirect.com/science/article/pii/S07475632140058
Measure Oct Nov Dec Jan 1st Jan 2nd 95
half half [4] K. S. Scott, K. H. Sorokti, and J. D. Merrell, “Learning “beyond the
Density 0.04 0.04 0.25 0.24 0.03
All Links 197 157 14 11 754 classroom” within an enterprise social network system,” The
Node Count 78 72 12 11 180 Internet and Higher Education, vol. 29, pp. 75 – 90, 2016. [Online].
Available:
http://www.sciencedirect.com/science/article/pii/S1096751615300
Network density refers to the measure of how the network 117
nodes connect to the other nodes. The measure of network density [5] H. El-Ghalayini, M. Abu-Arqoub, G. Issa, and A. F. Shubita,
is one when each node connects to every other node in the network. “Graduation-project management system: A social network analysis
perspective,” Journal of Theoretical and Applied Information
The highest network density value is from December 2019 to the Technology, vol. 95, no. 4, pp. 810 – 819, 2017. [Online]. Available:
first half of January 2020. During these times, fewer faculty http://www.jatit.org/volumes/Vol95No4/9Vol95No4.pdf
members and students have interacted with the system. Thus, the [6] O. E. Llantos and M. R. J. E. Estuar, “Characterizing instructional
hierarchical structure is relaxed, providing more connectivity to leader interactions in a social learning management system using
the nodes. However, when hierarchical structure emerges as in the social network analysis,” Procedia Computer Science, vol. 160, pp.
149 – 156, 2019.
interaction of the principal in the second half of January, network [7] O. E. Llantos and M. R. J. E. Estuar, “my.eskwela: Designing an
density is at its lowest. The low network density value means that enterprise learning management system to increase social network
more nodes do not connect to other nodes like the principal cannot and reduce cognitive load,” Procedia Computer Science, vol. 138, pp.
be directly related to the students and parents, and not all students 595 – 602, 2018.
relate to all the teachers in the school. [8] O. E. Llantos, “Cloudification of my.eskwela for e-governance in
philippine education,” Procedia Computer Science, vol. 109, pp. 680
– 685, 2017.
IV. CONCLUSION
The evolution of the sLMS social network emphasizes
engagement as the primary driver for network connectedness.
Engagement also affects to the centrality score of the roles. Thus,
for a social network with imposed hierarchies, critical roles need to
engage in the network to encourage roles below the hierarchy to do
the same.
138
Manufacturing and
Machinery
Session Chair : Dr. Arthur L. Fajardo
Moderator : Dr. Napoleon Enteria
Panelists : Dr. Sherwin Guirnaldo
Dr. Carl Salaan
Dr. Omar F. Zubia
Development of an Oyster Mushroom (Pleurotus
Spp.) Substrate Compactor-Bagger
Jordan L. Abad1,a *, Helen F. Gavino2,b, Romeo B. Gavino2,c and Melba M. Denson2
1
Department of Science and Technology Regional Office 1, San Fernando City, La Union
2
Department of Agricultural and Biosystems Engineering, College of Engineering, Central Luzon State University,
Science City of Munoz, Nueva Ecija
a
jlabad@region1.dost.gov.ph
b
hfgavino0313@gmail.com, caeigsstudent@gmail.com
*
corresponding author
Abstract— The low supply of mushroom in the country livestock and forestry products and cultivates major crops like rice,
evident to 90% importation is ironic to the abundant biomass coconut, and sugarcane, the state can generate substantial amount of
resource the country could offer. The low biomass utilization rate residues favorable for the production of mushrooms. However, the
affected by the lack of appropriate technologies also applies to the Philippine mushroom industry set its lowest production volume of
low productivity in bagging of substrate in the mushroom industry 355 metric tons (MT) in 2009 and worsened since 1995 [2] as almost
using manual labor with a capacity of 25 bags/hr. These problems 90% of the mushroom consumed was imported from different
enthuse the study to design, fabricate and evaluate a mushroom countries of South East Asia [3].
substrate compactor-bagger adoptable to local practices and The lack of appropriate research for mushroom
mushroom production technologies in the country. mechanization facing the country's small-scale and or village-level
The 1432mm x 741mm x 2040mm (l x w x h) 140kg mushroom industry could be observed in the bagging of substrate
machine was designed using 152 x 305mm, 0.02 mil polypropylene materials where manual labor is employed and described to be labor-
plastic bags with compaction height of 203mm and 95mm diameter. intensive [4]. As [5] explained it, more people are needed to do the
A weight of 750 grams of substrate per bag was determined as filling of bags than in other operations before bagging.
setting using the CLSU formulation (70-30 RS-S) and BSU (80-15- Mechanization technology suited for the scale of
5 S-RB-L) formulation. production in the bagging or filling of substrate bags and other
Performance evaluation of the machine using the two related operations could encourage the use of available biomass
formulations was tested with different loading weights of 10kg, resource and could further boost the mushroom industry towards
15kg and 20kg per batch. It was evaluated using simple CRD and self-sufficiency of mushroom supply in the country.
means were compared using LSD test. Results of the study revealed With these problems of low mushroom supply and the
that CLSU formulation at 15kg loading gained 71.67% bagging potentials of untapped biomass resource in the country, a need for
efficiency, 70.78% filling efficiency, 204 bags/hr bagging capacity efficient and cost-effective biomass utilization technologies and
and with an average of 740.90 grams/bag. Using BSU formulation, equipment compared to existing production technologies could be
the machine loaded with 20kg produced the highest bagging and seen to encourage the cultivation of mushroom. Thus, the study
compacting efficiencies of 88% and 83.48%. It had a capacity of 307 aimed to develop an oyster mushroom (Pleurotus spp.) substrate
bags/hr with an average of 759 grams/bag. The average weight per compactor-bagger using locally available materials and fabrication
bag was within the acceptable weight of 750 ±50 grams. technologies. The specific objectives of the study include the
The cost analysis of the machine for a year of operation at establishment of the optimum weight of substrate in bags; to design
six hours per day, three times a week with a custom rate of a substrate bagger adaptable to existing technologies of mushroom
PhP1.5/bag could generate an income of PhP96,736.00- production in the Philippines; to fabricate the machine using local
PhP191,749.00 per year and a payback period of 0.30-0.60 years for manufacturing technologies; to evaluate the operating performance
the BSU and CLSU formulations, respectively. of the machine in terms of capacity, efficiency, and uniformity in
With the developed machine, there would be increased weight of substrate bags; and, to perform cost analysis in using the
biomass waste utilization, more production of oyster mushroom at machine.
lower cost and hence, more benefits to the farmers and the With the study, bagging of the mushroom substrate would
mushroom industry.
be made easier resulting in higher output in terms of mushroom
Keywords— biomass utilization, compactor-bagger, substrate in bags.
oyster mushroom, substrate, formulation The creation of the machine could also lead to increasing
biomass resource utilization which was not given too much
I. INTRODUCTION attention. It would also provide opportunity to the local machinery
industry as this machine could be added to the list of machines that
According to economic analysts, mushrooms are seen as could be fabricated locally, thereby may also increase the income of
another significant crop of the country due to its growing local local machinery shops.
demand [1]. Being an agricultural country with well-endowed Thus, the study would help improve the production of the
mushroom substrate and the mushroom itself creating a more
140
profitable farmer but also in the utilization of biomass resource for where: 𝑉ℎ = 𝑣𝑜𝑙𝑢𝑚𝑒 𝑜𝑓 ℎ𝑜𝑝𝑝𝑒𝑟, 𝑚3
food production and profit creation. 𝑊1 = 𝑤𝑒𝑖𝑔ℎ𝑡 𝑜𝑓 𝑠𝑢𝑏𝑠𝑡𝑟𝑎𝑡𝑒 𝑝𝑒𝑟 𝑏𝑎𝑡𝑐ℎ, 𝑘𝑔
𝜌 = 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 𝑜𝑓 𝑠𝑢𝑏𝑠𝑡𝑟𝑎𝑡𝑒, 𝑘𝑔/𝑚3
II. METHODOLOGY
The design of the spiral conveyor was based from [6] for
The study was set to design, fabricate, and evaluate a
screw conveyors. The diameter of the shaft was calculated using
machine mechanizing the compacting and bagging of mushroom
Equation 2 where the density of sawdust was used, and the diameter
fruiting bags using two of the most commonly used formulation, the
of the spiral conveyor was 4”.
(a) 70 parts rice straw and 30 parts sawdust or CLSU Technology
and (b) 80-15-5 parts of sawdust-rice bran-lime or BSU
CT = (D2-d2)/36.6) x P x RPM (2)
Technology.
The study started with the gathering of benchmark data to where: CT = theoretical capacity (x 60%), ft3/hr
support the claim of mechanizing the bagging of the mushroom D = spiral diameter, in
substrate. The data gathered in the form of a survey questionnaire d = shaft diameter, in
was conducted related to the existing technologies in the country on P = pitch diameter, in (usually equals to D)
bag cultivation. Data gathered like the dimension of polypropylene RPM = speed of shaft, rpm
plastic (PP) bag, the capacity of laborer and weight of substrate
served as basis or design considerations for the fabrication of the The size of the PP bags dictated that the diameter of the
machine. compression chamber was 95mm and the height was 295mm lower
than the computed value of 305mm to facilitate easier mounting of
A. Establishment of Optimum Weight of Substrate in Bags PP bag wherein the remaining gap would be augmented by using
To standardize the weight of the substrate bagged by the the adjustable stopper depending on the type of substrate.
machine, the establishment of the optimum weight of substrate in After the dimensions and sizes of components were
bags was conducted. It was used in the conceptualization of some calculated, a final design plan was drawn using SolidWorks
settings for the fabrication of the machine. software.
The planting spawn weighs 40 grams per bag placed in
102mm x 152mm PP bags plugged with cotton through a PVC neck. C. Fabrication and Assembly of the Machine
It was sterilized in an autoclave for one and a half hour at 120 Local technologies and skills of manufacturing industry
degrees Celsius, 21psi. in the country was utilized in the fabrication using mainly the
Formulation of the substrate follows CLSU technology instruments found in a local fabricating shop. Some of the
and BSU technology. The substrate was manually bagged using equipment used were welding machine, grinding machine, drilling
152mm x 305mm PP bags with 0.02 mil thickness. The bags were machine, oxy-acetylene gas, and cutting disc. The assembled
approximately uniform in volume since height was maintained at components were set to their proper timing before conducting the
203mm while diameter at 95mm. It was inoculated with the prepared preliminary test.
spawn and was incubated in a dark room. It was then transferred to
the fruiting area after full ramification. D. Machine Testing and Evaluation
Data gathered include the number of days for full mycelial After the machine was fabricated, initial testing was
ramification, yield per bag, and flushing interval. Meanwhile, the conducted to validate the timing and functions of all components.
statistical analysis used was (2x3) factorial by Completely The study was laid out in Completely Randomized
Randomized Design. Factor A of the study included the types of Design, ANOVA was used to evaluate the effects of different
substrate mixture: (F1) 70% Rice Straw and 30% sawdust; and (F2) weights per batch in the machine using two formulations. Since the
80% sawdust, 15% rice bran, 5% lime while Factor B was the two formulations used differed in physical property especially the
different weight per substrate bag: (A) 500 grams, (B) 750 grams, bulk density of uncompacted materials, a separate evaluation of the
and (C) 1000 grams compacted in the same volume (95mm diameter machine using the two substrates was conducted. There were three
x 203mm height). varying weight per batch per kind of formulation. With this, there
were three replications per treatment for each of the two
B. Design of the Mushroom Substrate Compactor-Bagger formulations.
The machine had the following components: the flywheel The methods of preparation for the substrate materials in
and compactor assembly, the conveyor-hopper assembly, the the testing for the machine were similar with the preparation of a
rotating disc assembly, the mechanism to rotate the rotating disc substrate in the establishment of optimum weight per substrate bag.
assembly, the frame assembly and the power transmission system. Data Gathered during the evaluation include productive
The hopper was designed to accommodate 32-35kg of time, unproductive time, total bagging time, number of bags within
substrate per batch. The volume of the hopper was determined using the acceptable limit, total number of bags produced, and total weight
Equation 1 using density of uncompacted substrate for BSU of loaded inputs or substrate.
formulation of 347kg/m3 and CLSU formulation with 320kg/m3 . The bagging capacity of the machine was determined
using the Equation 3:
W1 (1) Bin
Vh = (3)
ρ C=
T
141
where: C= Bagging Capacity (bags/hr) To conduct the cost analysis of using the machine,
𝐵𝑖𝑛 = Number of Bags Produced within the limit (bags) assumptions were laid to facilitate computation. Six hours were
T = Total bagging time (hr) allotted for the total operating hours per day of the machine,
considering 8-hr operation for the laborer. The labor cost per laborer
The bagging and filling efficiencies of the machine were is PhP300.00/day at 6hr/day.
calculated using Equations 4 and 5. Assuming that in a community with several mushroom
and fruiting bag producers, 3 days/week was allotted for custom
Bin (4) hiring at PhP1.50/ bag. The custom rate is 75% of the manual
EffB = x 100
BT bagging cost of PhP2.00 considering the cost of manual bagging for
25 bags/hour. If the machine is used for a year, it would operate for
(5)
Win 156 days or 939 hours.
EffF = x 100 Considering that the machine uses 0.368kW of motor at
WT
7.13 pesos per kilowatt hour rate, the cost per hour of operation is
where: 𝐸𝑓𝑓𝐵 = 𝐵𝑎𝑔𝑔𝑖𝑛𝑔 𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 PhP2.64.
𝐵𝑖𝑛 = Number of Bags Produced within the limit (bags)
𝐵𝑇 = 𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑏𝑎𝑔𝑠 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑
𝑊𝑖𝑛 = 𝑇𝑜𝑡𝑎𝑙 𝑤𝑒𝑖𝑔ℎ𝑡 𝑜𝑓 𝐵𝑎𝑔𝑠 𝑊𝑖𝑡ℎ𝑖𝑛 𝑡ℎ𝑒 𝑙𝑖𝑚𝑖𝑡 (𝑔𝑟𝑎𝑚𝑠) III. RESULTS AND DISCUSSION
𝑊𝑇 = 𝑇𝑜𝑡𝑎𝑙 𝑊𝑒𝑖𝑔ℎ𝑡 𝑜𝑓 𝑙𝑜𝑎𝑑 𝑖𝑛𝑝𝑢𝑡 (𝑔𝑟𝑎𝑚𝑠)
A. Optimum Weight of Substrates
E. Cost Analysis For the average mycelial ramification, the analysis of
To evaluate the cost of using the machine, the following variance revealed significant difference between the formulation,
equations by [7] as presented in Table 1 were used. weight of substrate, and their interactions. The comparison of means
revealed that the 500g weight per bag with 26 days of full
Table 1. Formula used for the simple cost analysis of the study. ramification vary significantly to that of the 750g and 1000g bags
PARTICULARS FORMULA with 31 days and 40.6 days. The mean ramification of 27.33 days
Salvage value, (10% IC) SV = 10% IC for CLSU formulation varied significantly to the 37.73 days of BSU
Depreciation, PhP/yr D= (IC-SV)/n formulation. For the interaction of the factors, Table 2 shows that
Interest on Investment, PhP/yr I= ((IC+SV)/2)i the least ramification days was recorded in CLSU formulation at
Fixed Cost, PhP/yr Fc = D+I+TIS 500g/bag with 20 days only while highest at 41.20 days in
Variable Cost, PhP/yr Vc = Cp+Cl+RM+Club 1000g/bag BSU formulation. The simple effects vary significantly
Total bagging cost, PhP/yr Tc = Fc+Vc with each other except to that of BSU at 750g and CLSU at 1000g
Bagging Cost, PhP/bag Bc = Vc/C*T with 40 days ramification period.
Breakeven point, bag/yr BEP = Fc/Cr-Bc
Net Income per year, PhP/yr NI = C*T*Cr-Tc Table 2. Average mycelial ramification (days) per formulation as
Payback Period, yr PBP = IC/IN affected by different weight of substrate.
WEIGHT, g MEAN
where: Fc = Annual fixed cost, PhP/yr FORMULATION
A-500g B-750g C-1000g
D = depreciation, PhP/yr
I = interest on investment, PhP/yr CLSU 20 z 22 y 40.00w 27.33b
TIS = tax, insurance and shelter, PhP/yr BSU 32 x 40 w 41.20v 37.73a
Vc = variable cost, PhP/yr MEAN 26l 31k 40.60j 32.53
Lu = lubricant cost, PhP/yr Means with the same letter are not significantly different
Cp = fuel cost, PhP/yr
RM = repair and maintenance cost, PhP/yr The results revealed that the higher the weight of substrate
Cl = labor cost, PhP/yr in bags the longer it would take for the mycelial run for both
Tc = annual cost, PhP/yr substrate formulations. This result is supported by a similar study on
C = Bagging capacity, bags/hr the mycelial run for Pleurotus florida by [8] which resulted in lesser
Vc = variable cost, PhP/yr time for colonization in rice straw-based substrate than the
Bc = Bagging Cost, PhP/bag traditional substrate of sawdust.
Cu = cost of using the machine, PhP/bag The results of the analysis of variance of the average yield
T = Annual operating Time, hrs/yr per bag as affected by substrate formulation and weight per bag
Cr = Custom Rate, PhP/bag show significant difference between the weight per bag and
Break-even point = the number of bags that the machine interaction while no significant difference in formulation. The
would properly compact and bag to recover the capital comparison among means as affected by weight per bag presented
investment in Table 3 would show that the 750g produced the average yield per
Payback period = the time it takes to recover the investment bag of 130.25g and was significantly different to the 113.65g/bag of
cost the 500g substrate bag and the 76.75g yield of the 1000g. For the
interactions, yield for both formulations at 500g and 750g did not
142
differ significantly with each other but differ significantly to that of use of the hanging method if hanged one after the other. Meanwhile,
the 1000g/bag. the heaviest weight recorded the highest span of ramification in
days, lowest yields and the highest flushing intervals. The
Table 3. Average yield (g) per formulation as affected by different recommended setting of 750 grams for both formulations has been
weight of substrate. seen to prove the not too tight not too loose recommendation of
WEIGHT, g MEAN various authors and books on mushroom cultivation.
FORMULATION
A-500g B-750g C-1000g
B. Description and Operation of the Machine
CLSU 133.50v 73.40w 105.37 The major components of the machine were the following: (1.)
109.21v
BSU 118.10v 127.00v 80.10w 108.40 spiral-conveyor and hopper assembly, (2.) the four-cylinder rotating
MEAN 113.65k 130.25j 76.75l 106.89 disc assembly, (3.) the power transmission system, (4.) the piston
compressor assembly, (5.) the frame assembly, (6.) push-to-rotate
Means with the same letter are not significantly different lever assembly and (7.) the molder or compaction chamber
assembly as shown in Figure 1. These components were fabricated
The result in yield was similar to the results of [9] on oyster and assembled to form an operational compactor-bagger machine.
mushroom yield using tray method of cultivation whereby from 2kg
to 10kg per tray, 6kg was seen to be favorable for growing. The
results emphasized that the decrease in yield below the optimum
weight was attributable to earlier depletion of nutrients due to
limited substrate quantity while increasing the substrate quality did
not increase in mushroom production but only a waste of resources.
The comparison among means in Table 4 shows that the
flushing interval means of BSU formulation with 11.40 days
significantly differed to CLSU with 18.07 days flushing interval. For
the weight of bags, 750g per bag resulted to lowest flushing interval
of 8.75 significantly different to the 14 days of 500g and 21.45 days
of the 1000g. The interaction effects of the two factors were seen to
be significantly different with each other where CLSU at 750g with
8 days flushing interval differ significantly to all other treatment
combinations except with BSU at 750g with 9.50 days. The BSU
formulation at 750g also did not vary significantly to BSU at 500g
with 11.10 days but differed to other treatment combinations.
Meanwhile, BSU formulation at 500g and 1000g did not differ
significantly with each other but differed from the rest of the
treatment combinations.
Figure 1. The mushroom bagger-compactor
Table 4. Average flushing interval (days) per formulation group
The machine was made of locally available materials and
as affected by different levels of weight.
was fabricated using local manufacturing technologies. It was
WEIGHT MEAN
FORMULATION powered using a 0.367kW electric motor and was equipped with
A-500g B-750g C-1000g wheels for mobility in the mushroom house.
CLSU 16.90 w 8.00 z 29.30 v 18.07a The machine was operated by three persons with a
BSU 11.10 xy 9.50 yz 13.60 x 11.40b specified task. One who supervised the level of the substrate in the
MEAN 14.00k 8.75l 21.45j 14.71 feeding hopper, the other in putting the PP bag in the molder, and
Means with the same letter are not significantly different the last person in charge of the exit of the substrate bags or the
collector of the compacted and bagged substrate.
Based on the results of the statistical analysis conducted,
both formulations were affected by the different weight levels of C. Performance Characteristics of the Machine Using CLSU
substrate in bags. In selecting what type of formulation would be Formulation
more favorable than the other, CLSU formulation revealed better The machine was evaluated in terms of bagging efficiency,
results in terms of mycelial ramification (days) and flushing interval filling efficiency and bagging capacity.
(days) than BSU mixture but had no difference in yield for two The average bagging efficiency of the machine as affected
flushing periods. by different weight per batch using the CLSU formulation is
For the weight of substrate used in both formulations, presented in Table 5. Analysis of variance for the bagging efficiency
although the 500g substrate recorded the least ramification span for revealed that the means of bagging efficiency as affected by the
both formulations, 750g had favorable results in terms of yield and weight per batch differ significantly with each other at 5% level of
flushing intervals. During the study, it was also observed that the significance.
500g substrate formulation was difficult to manage as they were Comparison among means using Least Significant
easily disintegrated and substrate materials would shrink with the Difference (LSD) at 5% level of significance shown in Table 5
143
revealed that weight per batch of 15kg with highest bagging The highest bagging efficiency was computed in the 20kg weight
efficiency vary significantly with the 10kg and 20kg weight per per batch with 88% decreasing as the weight per batch decrease with
batch. 15kg in 84.21% and 10kg at 83.33%. Analysis of variance shows
that the weight per batch had no significant effect in the bagging
Table 5. Average bagging efficiency (%) as affected by different efficiency of the machine when using the BSU formulation at 5%
weight per batch. level of significance.
WEIGHT PER REPLICATE MEAN
BATCH, kg I II III Table 8. Average bagging efficiency (%) at different weight per
10 46.15 38.46 53.85 46.15 b batch.
15 75.00 75.00 65.00 71.67 a WEIGHT PER REPLICATE MEAN
20 2.63 17.65 21.88 14.05 c BATCH, kg I II III
Means with the same letter are not significantly different 10kg 83.33 83.33 93.33 83.33
15kg 84.21 84.21 84.21 84.21
It is also noted that using the weight per batch type, the 20kg 88 88 88 88.00
machine produced acceptable weight per bag ranging from 708.13
to 740.90g/bag and with an average of 725.42g/bag when using the Also, the bagged substrate using the BSU formulation
CLSU formulation. resulted to uniform weight per bag of substrate within the acceptable
On the filling efficiency of the machine as affected by limit with an average of 755.25g/bag close to the 750g/bag.
weight per batch using CLSU formulation, analysis of variance Meanwhile, results of the analysis of variance for filling
showed that there was a significant difference in the mean effects efficiency revealed that the treatment means varied significantly
of weight per batch in the filling efficiency of the machine at 5% with each other. Comparison among means using LSD as presented
level of significance. in Table 9 shows that the three weight per batch differed
Comparison among means presented in Table 6 revealed significantly with each other with 20kg as the highest filling
that weight per batch of 15kg with 70.78% vary significantly with efficiency of 83.48%.
the 10kg and 20kg weight per batch. The lowest recorded filling
efficiency is at 20kg weight per batch with only 16.75%. Table 9. Average filling efficiency (%) as affected by different
weight per batch.
Table 6. Average filling efficiency (%) as affected by different Weight per batch, kg REPLICATE MEAN
weight per batch. I II III
WEIGHT PER REPLICATE MEAN 10 75.90 75.90 75.80 75.87c
BATCH, kg I II III 15 79.60 79.60 80.20 79.80b
10 42.40 35.6 49.40 42.47 b 20 84.00 83.35 83.10 83.48a
15 73.60 74.13 64.60 70.78 a Means with the same letter are not significantly different
20 3.75 21.60 24.91 16.75 c
Means with the same letter are not significantly different Based on Table 10, the bagging capacity as affected by
weight per batch in the BSU formulation recorded 306.59~ 307
The summary of the machine’s bagging capacity as bags/hr when using the 20kg weight per batch. It was followed by
affected by different weight per batch is presented in Table 7. The the 15kg weight per batch of 302.11 bags/hr and 10kg with 292
analysis of variance of the bagging capacity affected by the weight bags/hr. Results of the analysis of variance revealed significant
per batch at 5% level of significance revealed no significant difference in the mean bagging capacity as affected by weight per
difference between means at 15kg and 10kg weight per batch but batch.
varied significantly with 20kg weight per batch. Comparison of means presented in Table 10 shows that
the bagging capacity in 20kg and 15kg weight per batch did not vary
Table 7. Average bagging capacity (bags/hr) as affected by significantly but differed significantly to the 10kg weight per batch.
different weights per batch.
WEIGHT PER REPLICATE MEAN Table 10. Average bagging capacity (bags/hr) as affected by
BATCH, kg different weights per batch
I II III
10 171.43 130.43 201.6 167.82 a WEIGHTS PER REPLICATE MEAN
a BATCH, kg I II III
15 214.29 215.14 180 203.14
b 10 300.00 288.00 288.00 292.00b
20 8.18 53.33 64.62 42.04 15 303.16 300.00 303.16 302.11a
Means with the same letter are not significantly different 20 306.98 304.62 308.17 306.59a
Means with the same letter are not significantly different
D. Performance Characteristics of the Machine Using BSU Other observations during the evaluation of the machine
Formulation include the elimination of PP bag breakage which was observed in
The average bagging efficiency of the machine as affected the manual bagging of substrates. This could be attributed to the
by weight per batch and BSU formulation is presented in Table 8.
144
controlled force of compaction as compared to manual labor with The cost curve of using the two formulations in the
varying forces exerted during bagging. machine is shown in Figure 2. The figure shows that the area
covered by the net income of the BSU formulation in using the
machine is greater than that of the CLSU formulation and that the
E. Cost of Using the Machine for the two types of Formulation custom rate is higher than the cost of producing per bag for the two
To determine whether the machine was cost-effective in formulations.
bagging substrates, a simple cost analysis was conducted.
The analysis used the highest capacity recorded for both
formulations as a result of evaluating the performance of the
machine using the treatments. For the CLSU formulation, 204
bags/hour were used as the bagging capacity as it showed significant
results based on the findings of the study while 307 bags/hour when
using the BSU formulation.
The machine was assumed to be operational thrice a week
with a total of 156 days/year or 939 hours per year total operating
time for every six-hour per day operation.
There were three operators for the machine with
PhP300.00 per laborer as labor cost. The custom rate of PhP1.5/bag
was based on the 75% cost per bag for manual bagging with
PhP2.00/bag pegged at PhP300.00 per day at 6-hour per day. It was
used since there was no existing bagger in the locality. While the
calculated values are presented in Table 11.
Table 11 shows that the total bagging cost per year in
using the machine is PhP190,731.92 with labor as the main
contributor with PhP140,785.71 for the three operators for a year.
Figure 2. The Cost Curve of the Machine using the two
The bagging cost for the CLSU formulation resulted to
formulations
0.995~1.00 peso per bag while 0.66 pesos per bag for BSU
formulation. The difference is due to the higher capacity of the later
contributing to a higher net income per year with a total of
PhP191,749.00. Meanwhile, the net income for CLSU formulation IV. CONCLUSION
is PhP96,736.00.
Taking into consideration the objectives of the study and
The computed breakeven point for the CLSU formulation
the conducted evaluation of the machine based on the parameters
is 32,997 bags per year with a projected payback period of 0.60 years
set, the following are therefore concluded:
of machine operation for the investment to be recovered.
When using the BSU formulation, 0.304 years is the
1. Compaction of substrate bags for CLSU formulation and BSU
payback period of the machine considering the assumptions. The
formulation using 152mm x 305mm PP bag is at optimum weight of
breakeven point for bags per year is calculated at 22,119 bags/year
750 grams/bag to attain optimum days of mycelial run, flushing
using the custom rate of PhP1.50/bag.
intervals and yield in oyster mushroom production;
Table 11. Comparison of cost analysis results for the two substrate
2. The designed compactor-bagger for oyster mushroom considers
formulations.
the mushroom technologies of the country to compact 750 grams of
PARTICULARS CLSU BSU substrate material in a 152mm x 305mm PP bag utilizing piston-type
Specifics of compression. It is designed to bag at least 100% greater than the
Salvage value, (10% IC) SV 5,825.7 5,825.7 manual bagging capacity of 25 bags/hr;
Depreciation, PhP/yr D 10,486.26 10,486.26
Interest on Investment, PhP/yr I 6,728.68 6,728.68 3. The fabricated machine could be fabricated using local
Taxes, Insurance and Shelter, TIS 2,912.85 2,912.85 manufacturing technologies and is made of locally available
PhP/yr (5% IC) materials. It is recommended for CLSU and BSU formulation;
Fixed Cost, PhP/yr Fc 20,127.79 20,127.79
Variable Cost, PhP/hr Vc 181.77 181.77 4. The capacities of the machine divided by three operators is still
Total bagging cost, PhP/yr Tc 190,731.92 190,731.92 more competitive than manual bagging capacity of 25 bags/hr with
Bagging Cost, PhP/bag Bc 0.995 0.66 68 bags/hour-102.33 bags/hour. The machine is also seen to
eliminate the breakage of plastic bags which was observed in manual
Breakeven point, bag/yr BEP 32,997.00 22,119.00
bagging;
Net Income per year, PhP/yr NI 96,736.00 191,749.00
Payback Period, yr PBP 0.60 0.304
5. The use of the machine could produce bags within the acceptable
limit of ±50 grams of 750 grams ranging from 740.90-758.94 grams
depending on the type of formulation used. The machine decreases
145
the gap of ununiformed weight per bags in manual bagging that [2] Chang, H. Y., Jeon, S. W., Cosadio, A. L., Icalina, C. L.,
could result to inconsistent days of ramification, yield, and flushing Panganiban, R., Quirino, R., & Song, Y. (2014). Status and
intervals; and, Prospect of Mushroom industry in the Philippines. Journal of
Philippine Association of Institutions for Research.
6. The cost of production per bag using manual labor is decreased [3] Business Diary Ph (2015, January 31). DA puts up P8 million
from PhP2.00 to PhP0.66-PhP0.995/bag. When the machine is used Mushroom Technology Center in Tarlac to boost production,
six-hour a day, thrice a week, at custom rate of PhP1.50/bag income substitute mushroom import from Taiwan. Retrieved from
Make Money – Business Diary Ph.:
to be generated may range from PhP96,736.00-191,749.00/year and
http://businessdiary.com.ph
a payback period of 0.304-0.60 years when using either CLSU and [4] De la Torre, E. (2010). Strengthening the Capabilities of
BSU formulation. climate change risk management and Disaster Preparedness in
Selected Provinces of the Philippines (Bicol Region). FAO,
ACKNOWLEDGEMENT
DA.
The support of DOST, the Department of Agricultural and [5] Oei, P. (2005). Small-scale mushroom. Digigrafi, Wageningen,
Biosystems Engineering of the College of Engineering through the The Netherlands: Agromisa Foundation and CTA.
project leader of the CLSU-DOST-ERDT Program Dr. Ireneo C. [6] Henderson, S. Milton., Perry, R. L. (1955). Agricultural
Agulto, and the Central Luzon State University- Center for Tropical process engineering. New York: Wiley.
Mushroom Research and Development (CLSU-CTMRD) with its [7] Hunt, D. O. (2001). Farm Power and Machinery Management.
Iowa State University Pres. Ames, Iowa. 10th Edition. Reissued
project leader- Dr. Sofronio Kalaw, and the Benguet State
2008 by Waveland Press Inc.
University -Mushroom Project is highly recognized in this paper. [8] Mondal, S., Rehana, J., Norman, M., and Adhikary, S. (2010).
Comparative study on growth and yield performance of oyster
REFERENCES mushroom on differerent substrates. journal of Bangladesh
Agriculture University.
[1] Philippine News Agency. (2010). Raise mushroom; DOST
[9] Zireva, D.T., M. Fanadzo and Mashingaidze, A.B. (2007).
urges farmers. Retrieved from www.balita.ph: Effect of Substrate Quantity and Shelf Position on Yield of
http://balita.ph/2010/07/06/raise-mushroom-dost-urges- Oyster Mushroom (Pleurotus sajor caju). Pakistan Journal of
farmers/ Biological Sciences, 10: 3458-3461.
146
Development of a Compost Tea Brewer Machine
Elman C. Torres1, a*, Emmanuel V. Sicat2, Marvin M. Cinense3 and Carolyn Grace G. Somera4
1,2,3 & 4
Central Luzon State University, Science City of Muñoz, Nueva Ecija
a
engr.ectorres@gmail.com
Abstract—This study was carried out to develop and require any special technology beyond a steeping vessel and is
evaluate the performance of the compost tea brewer machine. The associated with low cost or low energy input, whereas ACT requires
specific goals were to solve the problem of the previous design of constant stirring and aerating of large volumes of liquid.
the compost tea brewer machine with regards to the inability to
maintain the aerobic condition of compost tea during brewing Aerated compost tea a more recent approach adopted by
process, inability to extract available nutrients and microorganisms compost tea brewers involves an active process that relies on the use
available in compost, and design problems which result to difficulty of an aerator to oxygenate the mixture of compost and water during
in properly cleaning the machine. The developed machine was the brewing process, thereby producing aerated compost tea in a
evaluated in terms of the quality of compost tea produced. Three shorter brewing time ranging from 12 hours to 72 hours.
settings of the rotational speed of the mixing vessel (1rpm, 3rpm,
and 5rpm) were evaluated on brewing compost and its influence on However, there are documented problems with the recent
the quality of compost tea produced. Data gathered were analyzed design of compost tea brewer machines available in the market. For
using Random Complete Block Design and Least Significant instance, the inability of several machines to achieve aerobic
Difference was used for comparison of treatment means. Blocking conditions (6mg/L of dissolved oxygen level and above), inability
was done on each batch of produced vermicompost used in the to extract available nutrients and microorganisms from the compost
study. Results showed that 5rpm setting of vessel mixing and design of the machine that leads to the inability to properly clean
vermicompost gave a good quality of produced compost tea the machine.
(dissolved oxygen of 6.43mg/L, electrical conductivity of 1.530mS-
cm-1, pH level of 8.27, total bacterial count of 7.02log10 CFU/mL, Locally available design of aerated compost tea brewer
total NPK content of 900mg/L and seed germination index of machine includes an air diffuser for increasing oxygen level and
133.97%) which is above the m values set in this study. Based on having a vessel to hold the compost during the brewing time. Aside
the compost tea brewer machine’s cost of Php 43,033.00 with a from the attached air diffuser and several openings for the
capacity of 100 Liters of compost tea per brewing process, cost interaction of water and compost on the vessel, there are no other
analysis indicated a breakeven of 45 brewing cycles per year and a mechanisms that will cause the agitation of compost inside the
payback period of 2.4 years. vessel which can lead to the inability of the machine to extract
available nutrients and microorganisms and it can lead to the
Keywords—Aerobic Condition, Aeration System, anaerobic condition of the compost. To overcome the constraints
Compost Tea Brewer Machine. discussed above, it is necessary to develop a new compost tea brewer
machine that will address the above problem using locally available
I. INTRODUCTION materials.
147
allowable stresses of the selected materials. The design drawing was E. Compost Tea Brewing Procedure and Data Gathering
drafted using the SolidWorks software. Further analyses of the
major elements were also done using such features of SolidWorks. Compost tea brewer machine with all necessary
instrument, apparatus, and brewing materials were placed in a
Based on the approved design plan, the machine was covered area with electrical power source. Brewing tank was filled
fabricated by an accredited local machinery fabricator that had with 100 liters of water. Power switch of the air pump was turned
simple and local manufacturing tools and equipment. Stainless steel on. Water inside the brewing tank was aerated for 2 hours to raise
materials and HDPE plastics were used in the components that are the DO level from 4mg/L to 8 to 12mg/L and to dissipate the
in contact with the compost tea. Simple manufacturing processes, chlorine from water. The vermicompost (6.83kg) was placed on
namely: measuring, cutting, joining of different parts by welding, compost vessel and placed in the machine. The machine was allowed
bolts and nuts, assembly of the different components, finishing and to brew compost tea for 24 hours. After 24 hours of brewing process,
painting were employed by the local manufacturer. the machine was switched off and the gathering of samples of
compost tea was done. The same procedure was done for each
setting of rotational speed.
B. Preliminary Testing of the Machine
During brewing process, DO level was recorded every six
Preliminary tests were conducted to determine the (6) hours using pen type dissolved oxygen meter from the start up to
working conditions of different components of the machine. the end of brewing process at the surface liquid in brewing tank.
Necessary adjustments and/or minor modifications were done until Data gathered was used to determine the trend of DO level over time
the machine satisfied the required rpm and DO level to be of brewing process. Electrical conductivity and pH level were
maintained. Visual inspection was made on parts of the machine. recorded using electrical conductivity meter and pH meter at the end
This also allowed the familiarization of the researcher with the of brewing process at the surface liquid in brewing tank. One (1)
machine. A 24-hour test run was conducted to ensure that the sample of 350 mL compost tea was collected at the discharge after
machine will be able to attain the required rotational speed and brewing process and was placed on unused distilled water container.
minimum level of dissolved oxygen during the performance Each sample taken was subjected to laboratory testing to determine
evaluation. the total percentage of NPK in compost tea.
148
on a prepared data sheet. Performance parameter that was used as B. The Principle of Operation
reference in the evaluation is shown in Table 3.
The main function of compost tea brewer machine is to
Table 3. Minimum compost tea quality requirements. extract microorganisms, nutrients and organic compounds available
on compost into a water solution. Compost vessel held compost and
Item Unit Values the lower portion of compost vessel was suspended into the tank
Dissolved Oxygen Level, [5] mg/L 6 or higher , with water throughout brewing process. A motor caused the compost
pH Level, [8] - 6.5 – 8.4 vessel to rotate through chain and sprocket connection to mix the
Electrical Conductivity, [6] & [9] mS-cm–1 1.1 or higher compost inside the vessel and to allow extraction of microorganism,
257mg/L or nutrients and organic compounds present in the compost. Air pump
NPK Content, [10] mg/L
higher directed the air through PVC pipes into the tank which stirred its
Log10 contents and kept the consistency of the compost tea relatively
Population of Bacteria, [11] 5.00 – 9.00 uniform. It provided oxygen to support the growth of microorganism
CFU/ml
Germination Index, [12] % 60 or higher in compost tea. Brewing tank held the compost tea throughout the
brewing process.
149
D. Results of Preliminary Testing compost and water with ratio of 1:10 were used to test and evaluate
the performance of compost tea brewer machine.
After fabrication of the machine, preliminary testing was
done to check if the machine worked according to its purpose.
During the first preliminary testing, rotational speed of compost
vessel was established through series of different number of teeth of
the sprocket and using motor controller that regulated the rotational
speed of motor. A power switch was also provided at the right side
of the device for safety purposes and ease of operation.
150
recorded at three (3) rotational speed were above the minimum value aerated compost tea did not vary greatly (5.07 – 5.16) as affected by
of 1.100mS-cm-1 in which had a greater plant growth response [9]. duration of brewing process.
Analysis of variance revealed a statistically significant indication at
α=0.01 to show that there was a highly significant difference in Total Bacterial Count
electrical conductivity among the three (3) rotational speeds of
compost vessel to mix the compost. Comparison of among means of Total bacterial count (log10 CFU/mL) of compost tea as
the electrical conductivity as influenced by rotational speed of influenced by varying rotational speed of compost vessel is shown
compost vessel revealed that electrical conductivity of compost tea in Table 9. Highest mean of total bacterial count of compost tea with
produced at 1rpm setting was significantly lower compared to the 7.13log10 CFU/mL was recorded at 3rpm setting while lowest mean
electrical conductivity of compost tea produced at 3rpm and 5rpm total bacterial count was recorded at 1rpm with 6.99log10 CFU/mL.
settings. The electrical conductivity of compost tea produced at The values of total bacterial count of compost tea produced at three
3rpm and 5rpm did not differ significantly. It could be concluded different settings were within the ideal range of total bacterial count
that 3rpm and 5rpm setting produced compost tea with higher of compost tea (5.00 to 9.00log10 CFU/mL). The values of total
electrical conductivity compared to 1rpm setting. bacterial count were also in agreement with the range of results (6.43
to 8.62log10 CFU/mL) of other study of compost tea [10].
Table 7. Electrical conductivity (mS-cm–1) of compost tea as
affected by rotational speed of compost vessel. Table 9. Total bacterial count (log 10 CFU/mL) of compost tea
Rotational Block as affected by rotational speed of compost vessel.
speed, Mean Rotational Block
I II III speed, Mean
RPM
I II III
1 1.490 1.467 1.448 1.470b RPM
3 1.544 1.500 1.500 1.510a 1 7.091 7.15 6.73 6.99
5 1.550 1.540 1.510 1.530a 3 7.180 6.92 7.29 7.13
Means not sharing same letter in a column differ significantly at 1% 5 7.120 6.96 6.98 7.02
by Least Significant Difference (LSD).
Analysis of variance revealed no significant difference on
The higher rotational speed of vessel mixing compost the total bacterial count of compost tea produced among the three
helped the extraction of available soluble salts in the compost. rotational speed of compost vessel. The mean total bacterial count
Source of soluble salts in composts such as carabao manure consists of compost tea were 6.99log10 CFU/mL, 7.13log10 CFU/mL and
of ammonium, nitrate and phosphate and other major soluble salts 7.02log10 CFU/mL for 1rpm, 3rpm and 5rpm respectively.
such as sodium, magnesium, calcium and sulfate. Reference [14] Therefore, the total bacterial count of compost tea produced at 1rpm,
suggested that the electrical conductivity of compost tea should not 3rpm and 5rpm were the same.
exceed to 2.00mS-cm–1 or if excessive soluble salts exist, plant
injury, such as reduced germination rates and leaf burning, or death This could be attributed also to the compost being used in
might occur. the study and level of DO level which the air flow rate given aeration
system was constant. Reference [5] suggested that if mixing of
pH Level compost was not adequate, the compost would compact and
eventually would go to anaerobic condition because it did not have
The pH levels of compost tea as influenced by varying aerated water moving through the compost. The brewing process in
rotational speed of compost vessel is shown in Table 8. Highest all experimental units were above the minimum aerobic range of DO
recorded mean of pH level of compost tea was produced at 1 RPM level of about 6mg/L and above.
setting with 8.4 while lowest recorded mean of pH level was
produced at 3 RPM setting. The values of pH level of compost tea Total NPK Content
produced at three rotational speed setting were within the
recommended range of pH level (6.50 – 8.40). The total NPK content (%) of compost tea as influenced
by varying rotational speed of compost vessel is shown in Table 10.
Table 8. pH level of compost tea as affected by rotational speed The highest mean of total NPK content of compost tea was produced
of compost vessel. at 5rpm setting while the lowest mean of compost tea was produced
Rotational Block at 1rpm setting. The values in the experimental unit were above the
speed, Mean baseline set in the minimum total NPK (257mg/L) content based on
I II III other published study of compost tea [6] & [10].
RPM
1 8.4 8.4 8.4 8.40
Table 10. Total NPK content (mg/L) of compost tea as affected
3 8.3 8.1 8.1 8.13
by rotational speed of compost vessel.
5 8.2 8.2 8.4 8.27
Rotational Block
speed, Mean
Analysis of variance showed that there was no significant I II III
RPM
effect of varying rotational speed of compost vessel in the pH level
of compost tea produced. The pH level of compost tea were 8.40, 1 800 640 400 613.33
8.13 and 8.27 for 1rpm, 3rpm and 5rpm respectively. Therefore, it 3 800 900 900 866.67
can be concluded that pH level of the produced compost tea is the 5 700 800 1200 900.00
same whether it was produced at 1rpm, 3rpm and 5rpm setting.
Analysis of variance revealed no significant difference on
The pH level across the experimental unit did not vary the total NPK content of compost tea produced at three (3) different
greatly due to the same compost materials being used in each rotational speed setting. The mean values of compost tea were
experimental unit of the study. In reference [10], the pH level of 613.33mg/L, 866.67mg/L and 900mg/L for 1rpm, 3rpm and 5rpm
151
setting respectively. Therefore, it could be concluded that total NPK compost tea in all three (3) rpm setting. Downward trend of DO level
content of the produced compost tea was the same whether it was and upward trend of TBC could be attributed to the growth on
produced at 1rpm, 3rpm or 5rpm setting. population of bacteria in the compost tea. As the population of
bacteria increased the demand for oxygen increased and as a result,
Despite the insignificant result of analysis of variance, it the DO level decreased in the compost tea. The critical part of
could be observed that the higher rpm would have higher NPK brewing process is when the DO level of compost tea reached below
content compared to lower rpm, but several authors suggested that 5mg/L, anaerobic condition ensued and materials that were toxic to
solubility of nutrients could also be attributed to the quality of plant growth were produced in the compost tea.
compost and different levels of concentration of nutrients present in
compost to water [15]. In addition, it was observed that the highest
and lowest total NPK content and electrical conductivity of
produced compost tea were at 5rpm and 1rpm respectively. It only
indicated that the higher the rotational speed of compost vessel the
more nutrients would be present in compost tea due to agitation and
mixing of compost.
152
Cost Analysis of the Machine machine could be evaluated using shorter duration of brewing
process. articles.
The cost of materials and fabrication of compost tea
brewer machine was Php 45,033.00 with expected life span of 5 ACKNOWLEDGEMENT
years. Considering all the assumptions made including the annual
used of machine per year of 98 brewing cycle per year and price per The Engineering Research and Development for
liter of 8.00 Php/Liter, cost analysis indicated a breakeven of 45 Technology (ERDT) Scholarship, Department of Agricultural and
brewing cycle per year and a payback period of 2.4 years. Figure 6 Biosystems Engineering, College of Engineering and Central Luzon
shows the breakeven graph analysis of using compost tea brewing State University are hereby acknowledged in this research.
machine. It is revealed that the compost tea brewer machine could
earn a net income when the machine operates more than 45 brewing
cycle per year. REFERENCES
[1] brackets [1]. The TNAU Agritech Portal. (2009). Organic
Farming: Compost, URL:
http://www.agritech.tnau.ac.in/org_farm/orgfarm_composti
ng.html
[2] Scheuerell, S.J. (2004). Compost Tea Production Practices,
Microbial Properties, and Plant Disease Suppression. Soil
and Compost Eco Biology, International Conference of Soil
and Compost Eco-Biology, Leon, Spain, pp. 1.
[3] Weltzien, H. C. (1991). Biocontrol of foliar fungal disease
with compost extracts, Microbial Ecology of Leaves.
Springer-Verlag, New York, p. 430-450.
[4] Scheurell, S.J., & Mahafee, W.F. (2002). Compost Tea
Principles and Prospects for Plant Disease Control. Compost
Science and Utilization, Volume 10, 2002, pp. 313-338
[5] Ingham E.R. (2005). The Compost Tea Brewing Manual.
Soil Food Web Incorporated. Corvellis Oregon 97333. Fifth
Fig. 6. Breakeven graph analysis of using compost tea brewer Edition, pp. 1-5.
machine [6] Pant, A., Radovich, T.J.K., Hue, N.V. & Arancon, N.Q.,
(2011). Effects of vermicompost tea (aqueous extract) on
pak-choi yield, quality, and on Soil biological properties.
Compost Sci. Util. 19 (4), pp. 279–292.
IV. CONCLUSION AND [7] Ndegwa PM, SA Thompson and KC Das, 1998. Effects of
RECOMMENDATION Stocking Density and Feeding Rate on Vermicomposting of
Biosolids. Bioresource Technology, 71: 5-12.
Based on the results of the study, the following conclusion [8] Ayers, R.S., & Westcot, D.W. (1976) Water Quality for
were drawn: A compost tea brewer machine with five (5) major Agriculture. Food and Agriculture Organization of United
components namely: compost vessel, brewing tank, aeration system, Nations. Rome, Italy 1976, pp 7
transmission system, and main frame was successfully fabricated [9] El-Haddad, M.E., Zayed, M.S., El-Sayed, G.A.M.,
following the design calculations and considerations. It operated on Hassanein, M.K. & El-Satar A.M.A. (2014). Evaluation of
the principle of extraction of beneficial organism and nutrients Compost, Vermicompost and their Teas Produced from Rice
through rotational motion of vessel containing compost. The Straw as Affected by Addition of Different Supplements.
machine with an overall dimension of 1,365mm x 665mm x Annals of Agricultural Science (2014) 59(2), pp. 243-251
1,370mm was completely fabricated using available materials from [10] Martin, C.C.G. (2014). Potential of Compost Tea for
local manufacturing shop. Suppressing Plant Diseases. CAB Reviews. 9(32), pp. 1-38.
[11] Ngampimol, H., & V. Kunathigan, (2008). The study of
The machine performed satisfactorily based on its shelf life for liquid biofertilizers from vegetable waste.
performance parameters in terms of quality of the produced compost Assumption Univ. J. Technol., 11, pp. 204–210
tea (dissolved oxygen level, electrical conductivity, pH level, total [12] Zucconi, F. and De Bertoldi, M. (1987) Compost
bacterial count and total NPK content). Generally, the compost tea Specifications for the Production and Characterization of
brewer machine best operated at operating speed of 5rpm with Compost from Municipal Solid Waste. In: De Bertoldi, M.,
dissolved oxygen level of 6.43mg/L, electrical conductivity of Ferranti, M.P., L′Hermite, M.P. and Zucconi, F., Eds.,
1.530mS-cm–1, pH level of 8.27, total bacterial count of 7.02 log10 Compost: Production, Quality and Use, Elsevier, London,
CFU/mL, total NPK content of 900mg/L and seed germination index pp. 276-295.
of 133.97%. The cost analysis on the use of the machine revealed [13] Western Sustainable Agriculture Research and Education,
that it is financially viable if it is used for selling produced compost (2017). Aerated Compost Tea: A Field Guide to Production
tea to the farmer. The cost of fabricating the device was Php Methods, Formulas and Application Protocols,
45,033.00. The average annual net income was 18,739.78Php/year, Albuquerque, New Mexico 87197, pp. 3-5
with a computed breakeven brewing cycle per year and payback [14] Gartley, K.L. (2011). Chapter 10: Recommended Methods
period of 45 brewing cycle/year and 2.4 years respectively. for Measuring Soluble Salts in Soils. Recommended Soil
Testing Procedures for the Northeastern United States.
Based on the results of the study, the following were USA, pp. 88-93
hereby recommended: The compost tea brewer machine should be [15] Bernal, M.P., Paredes C., Sanchez-Monedero, M.A.,
used at operating speed of 5rpm and brewing time of 24 hours. The Cegarra, J ., Maturity and stability parameters of composts
machine could be evaluated using other types of compost. The prepared with a wide range of organic wastes, Bioresource
Technology, 63 (1998), pp. 91-99
153
Development of an Impact-type Cacao
(Theobroma cacao L.) Bean Huller
Herald Kam G. Hernando1,a*, Carolyn Grace G. Somera2,b, Jeffrey A. Lavarias2,c, and Emmanuel V. Sicat2,d.
Department of Agricultural and Biosystems Engineering, College of Engineering, Central Luzon State
University, Science City of Munoz, Nueva Ecija
heraldkam@gmail.com1,a*, cggsomera@clsu.edu.ph2,b, jeffreylavarias@clsu.edu.ph2,c,
emmanuelsicat@yahoo.com2,d
Abstract - The study mainly aimed to develop an impact type screw/auger, serrated cone, and hammer types that cause more
cacao bean hulling machine and evaluate its performance in breakage thus developing more dust and smaller particles in
producing cacao nibs in terms of input capacity, hulling the operation which decreases the recovery of large nibs with
capacity, hulling efficiency, nib recovery, large nib recovery, above 3mm particle size. Hence, the availability of a food
purity, noise level and power consumption. The cacao huller grade compliant, efficient impact-type cacao hulling machine
was subjected to three varying tangential velocities in may address the aforementioned problem.
cracking mechanism, specifically at 40m/s, 47m/s, and 54m/s,
and with three different volumes of air in winnowing II. METHODOLOGY
mechanism with 41m3/hr, 45m3/hr, and 49m3/hr. The data
gathered were arranged in a CRD Two Factorial and A. Preparation of Samples
comparison among means using Statistical Tool for Locally available dried cacao beans were used as
Agricultural Research (STAR) software. The analysis showed samples for the testing and evaluation of hulling machine.
that tangential speed of the machine significantly affected all Three kilograms of cacao beans in 27 trials was used in final
parameters except purity and noise level; moreover, the testing the machine. A food grade material was used to roast
volume of air significantly affected the hulling efficiency, nib the cacao beans until the beans started to crack (a cracking
recovery, large nib recovery and purity. Impact-type cacao sound is usually heard) and the outer shell became darker in
bean huller with 47m/s and 41m3/hr gave the highest input color. Moisture content of the roasted cacao bean was
capacity and hulling capacity of 195.60kg/hr and 156.53kg/hr, recorded and not longer an hour to be fed into the impact-type
respectively. Furthermore, the highest hulling efficiency, nib cacao bean huller due to cooling that may also lead to
recovery, and large nib recovery of 96.84%, 96.98%, and absorbing moisture.
86.75%, respectively, were obtained from tangential velocity
of 40m/s and volume of air of 41m3/hr. The lowest power B. Fabrication of the Machine
consumption and noise level were obtained from the The proposed mechanical cacao bean huller was
combination of 40m/s and 45m3/hr with 3.99kwh and fabricated by a certified agricultural machinery manufacturer
88.20dB, respectively. When considering mass production of in accordance to the approved drawings and design
cacao nibs, the most desirable set up of the machine was the specifications. Locally available materials were used in the
combination of 40m/s tangential velocity and 41m3/hr volume fabrication of the machine. The machine underwent
of air. preliminary testing to determine if the machine worked
Key Words - cacao beans, cacao nibs, hulling, nib according to the design purpose and if it satisfied the set
recovery, winnowing. criteria and to assess the working conditions and possible
operation faults of each component. Visual observations were
I. INTRODUCTION made to assess the safety, proper alignment, and actual
clearances that may affect the operation of the machine. The
Mechanization across the different areas of the final testing was conducted after the machine was desirable
value chain is necessary in gaining competitive and relative for its intended function and ready for performance
advantage in cacao industry. Improvement of postharvest evaluation. Running-in test was also conducted on an area
facilities is the primary challenge in the Philippine cacao with stable and level ground and adequate working space.
industry given its growing demand particularly for the high
value added products like cacao nibs, tablea, cacao powder C. Data Collection
and cacao butter. On cacao production and management, During the test and evaluation, the following data
majority of local farmers and cacao traders do not go beyond were gathered for analysis and determination of the
selling beans in the market due to high investments incurred performance parameters of the machine namely; moisture
in the hulling operation of cocoa beans. Cacao farmers and content, initial and final weight, weight of large nib, duration
traders are still using the conventional way of cracking and of test, noise level, speed of rotating components, air velocity,
winnowing roasted cacao beans to produce nibs. Cracking the and power consumption.
cacao beans is carried out to produce nibs using a roller pin
that is mostly used in bakeries. Meanwhile, winnowing D. Statistical Analysis
cracked cacao beans is accomplished by tossing the beans and The machine was evaluated based on the three
shells up into the air on a windy day using a winnowing tangential velocities of the rotating cracking mechanism and
basket also known as “bilao” which takes an hour to finish three different volumes of air in the winnowing system. It was
three to five kilograms of roasted cacao beans. As a result, replicated three times. Each experimental unit used 3 kg of
most local farmers and traders are not able to produce enough roasted cacao beans. The speed of the rotating mechanism was
nibs to supply the local chocolate manufacturers. Existing based on the resulting optimum speed during the preliminary
machines in the Philippine cacao industry employ cracking testing of the machine which was 2400rpm. The Analysis of
operation only wherein winnowing is done separately using Variance was also computed using STAR software and
conventional way. Moreover, existing design of cacao hulling Comparison among Means was done to parameter means with
machines uses multiple contacts to the beans including roller, significant effects.
154
III. RESULTS AND DISCUSSION 93.33%. Additionally, no significant effect was recorded in
noise level of the machine; however, tangential velocity of the
A. Description of the Machine cracking system significantly affected the power consumption
The impact-type cacao bean hulling machine is of the machine at 5% level of significance with 4.42kwh as
composed of hopper; impeller type cracking component; the highest record in the combination of 47m/s (tangential
winnowing component with aspirator and a cyclone that velocity) and 49m3 /hr(volume of air).
separates the cacao nibs to hulls and other impurities;
transmission system that consists of v-belts and pulleys which IV. CONCLUSIONS
was coupled to the main shaft powered by an electric motor;
nib and hull discharge; and the support frame. The fabricated Based on the result and observations of the study, the
impact-type cacao bean hulling machine as shown in Figure 1, following conclusions were drawn:
functions in a continuous manner wherein the material is a. The impact-type cacao bean hulling machine with
constantly fed into the hopper. combination of cracking and winnowing systems was
The hopper could carry a successfully designed based on different considerations;
capacity of 15 kgs of roasted b. The impact-type cacao bean hulling machine was
cacao beans when full. An
fabricated using locally available materials and local
impeller type cracking
chamber receives the input manufacturing technologies and functions in a continuous
material and is then passed to manner wherein the material is constantly fed into the hopper
the aspirator with a cyclone that could carry a capacity of 15 kgs of roasted cacao beans
type vacuum which separates when full;
the cracked beans to its hulls. c. The machine performed satisfactorily with highest
The cacao hulls are collected in input capacity and hulling capacity using the combination of
the cyclone while the cacao
Figure1 .Cacao Bean Huller 40m/s and 41m3/hr. Moreover, the combination of 40m/s and
nibs are then collected with a
pail supported by inclined nib outlet. The machine can be 49m3/hr obtained the highest hulling efficiency, nib recovery,
operated continuously by one person. All materials used in the and large nib recovery. Additionally, lowest power
design and fabrication were of food grade quality except the consumption and noise level of were recorded from the
support frame and cover, most of which was stainless steel combination of 40m/s and 45m3/hr. highest purity of roasted
especially components that are in direct contact of the roasted cacao nibs was recorded from the combination of 47m/s and
cacao beans. 49m3/hr. The machine delivered the most desirable
performance when set at 40 m/s and 41m3/hr.
B. Performance Evaluation
HIGHEST LOWEST
PARAMETERS ACKNOWLEDGEMENTS
VALUE VALUE
The Department of Science and Technology through
Input Capacity 195.60 kg/hr 166.39 kg/hr
Engineering Research and Development for Technology
Output Capacity 169.69 kg/hr 136.43 kg/hr
(DOST-ERDT) is highly recognized in this paper.
Hulling capacity 156.53 kg/hr 124.99 kg/hr
Hulling Efficiency 96.84%, 76.86% REFERENCES
Hulling Recovery 96.98% 45.29% [1] Adejuyigbe, S. B. & Bolaji, B. O. (2005). Design
Large Nib Recovery 86.75% 69.36% fabrication and performance evaluation of beans
Purity of Nibs 99.00% 93.33% dehuller. Journal of Science and Technology. Volume
Noise Level 91.17 dB 88.20% 25, No.1. http//:doi.org//10.4314/just.v25i1.32938
Power Consumption 4.42 kwh 3.99 kwh [2] Akinnuli, B. O., S. P. Ayodeji & Omieza, A. J. (2014).
Computer aided design for cocoa beans processing yield
Statistical analysis showed that tangential velocity of the prediction. International Journal of Applied Science and
cracking mechanism significantly affected input capacity and Technology, Vol. 4, No. 5.
hulling capacity of the machine at 5% level of significance. [3] Akinnuli, B. O., O. S. Bekunmi, & Osueke, C. O. (2015).
Moreover, the highest input capacity (195.60kg/hr) and Design concepts towards cocoa winnowing
hulling capacity (156.53kg/hr) of the machine were generated mechanization for nibs production in manufacturing
from the tangential velocities and volumes of air of 47 industries. British Journal of Applied Science and
m/sand41m3 /hr, respectively. Also, tangential velocities and Technology 8(1):35-45, 2015, Article no.
volumes of air of40m/sand 49m3 /hr garnered the lowest input BJAST.2015.184.
capacity and hulling capacities of 166.39kg/hr and http//:doi.org//10.9734/BJAST/2015/16161
124.99kg/hr, respectively. [4] Bureau of Plant Industry – Department of Agriculture
For the winnowing mechanism, statistical analysis (2017). Philippine cacao industry road map 2017-2022.
revealed that volume of air significantly affected the hulling [5] Lavarias, J. A. (2015). Optimization of mechanical cacao
efficiency, nib recovery, and large nib recovery at 5% level of shelling parameters using unroasted cocoa beans.
significance. The combination of 40m/s (tangential velocity) International Journal of Agriculture and Biosystems
and 41m3 /hr (volume of air) of the machine generated the Engineering. Vol:9, No:6, 2015.
highest hulling efficiency, nib recovery, and large nib waste.org/abstracts/29726
recoveries of 96.84%, 96.98%, and 86.75%, respectively. [6] Philippine National Standard (PNS) – Bureau of
The interaction of the two treatments significantly Agriculture and Fisheries Standards (BAFS) for
affected large nib recovery and purity of the roasted cacao Agricultural Machinery (2018). Cacao Huller
nibs. The combination of 47m/s (tangential velocity) and Specifications and Methods of Test.
49m3 /hr (volume of air) gained the highest purity of 99.00%,
while 54m/s and 41m3 /hr garnered the lowest nib purity of
155
Development of Dry-Scale Cleaner cum Grader for
Stored Onion (Allium cepa L.)
Jamel M. Salo1, a *, Emmanuel V. Sicat2,b, Helen F. Gavino3,c and Theody B. Sayco4,d
1
Carmen, Bohol
2
Science City of Muñoz, Nueva Ecija
3
Science City of Muñoz, Nueva Ecija
4
Science City of Muñoz, Nueva Ecija
a
jamsalo565@gmail.com, bemmanuelsicat@yahoo.com, chfgavino0313@gmail.com, dtbsayco@clsu.edu.ph
*graduate student Central Luzon State University
Abstract— The study was conducted to develop a dry- in each elements. A detailed drawing was used as a guide in the
scale cleaner cum grader for stored onion bulbs. The machine was fabrication of the machine at an accredited local fabricator in San
tested by using three levels of speed in the cleaning and grading Rafael Village, Navotas City. Modification on the major parts was
assembly (e.g. 53rpm & 13rpm, 61rpm & 15rpm and 71rpm done until its functionality and design considerations was achieved.
&17rpm) as factor. Based on the results, the most desirable machine
set up was established at cleaning speed of 71rpm and grading speed
of 17rpm. Evaluated performance parameters was cleaning capacity B. Performance Evaluation
(535.28kg/hr), grading capacity (556.53kg/hr), overall capacity Samples used were un-cleaned and ungraded onion bulbs
(493.29kg/hr), cleaning efficiency (88.50%), grading efficiency with visible loose dry scales due to curing and storing. Each sample
(82.20%), overall efficiency (72.71%), mechanical damage (2.09%), (10kg each) was randomly selected considering that no rotten bulbs
power consumption (14.41Wh), and noise level (82.51dB). The total were included. Machine performance was evaluated based on 3
machine cost was ₱125,000.00 with an operating cost of ₱0.40/kg. levels of speed combinations in the cleaning and grading assembly
It will break-even at 33,262.16kg at a custom rate of ₱1/kg in a year. (e.g. 53rpm & 13rpm; 61rpm & 15rpm; and 71 rpm & 17rpm) and
Based on the results, the machine can lessen the manual labour was replicated 3 times. Raw data such as initial weight, operating
utilization from 0.03 man-day/kg to 0.0008 man-machine-day/kg. time, impurities left and removed, weight in each outlet chutes,
weight of correct and correctly graded, weight of mechanically
Keywords— dry-scale cleaner cum grader, stored onion damage, power consumption and noise emitted was collected to
bulbs, desirable machine set up, performance parameters determine and analyze the performance parameters such as capacity,
efficiency, mechanically damage, power consumption and noise
I. INTRODUCTION level. The economic performance such as unit cost of operation,
break-even cost, and payback period were also determined.
Onion (Red Creole and Yellow Granex) is widely
cultivated in the Philippines with an earnings of P5.5B in which
Central Luzon is the major producer [1]. However, farmers have a C. Performance Analysis
difficult time competing to imported onions in the market [2]. As a Performance evaluation were set up according to
conventional practice, onion bulbs are stored to preserve its quality Complete Randomized Design (CRD). Analysis of Variance
and acquire higher price during off seasons [3]. However, it requires (ANOVA) was used to indicate statistical significance among
cleaning of excess dry-scale and grading base on its size to increase treatments and Post- hoc test using Fisher’s Least Significant
its marketability [4]. At present, manual cleaning and grading are Difference (LSD) was conducted to test which speed variations
still traditionally practiced. This processes requires an average labor shows significant results. The establishment of the desirable
utilization of 0.03man-day/kg [5]. Resulting to low output capacity, machine set up was based on simple rating system. The machine’s
inconsistency and non-availability of labors during peak seasons [6]. performance in three speed variations were rated in a scale of 1 to 3,
The study can help simplify and hasten the onion cleaning and 1 being the highest and 3 being the lowest based on the results of the
grading. It can also help to improve the uniformity, adds value, statistical analysis. The ratings were averaged and the result that is
lessen the man-day requirement and cost of operation. Hence, there closer to 1 was the desirable set up of the machine[8].
is a need to develop a machine that can clean by removing the excess
dry scales after storage and efficiently grade them according to size.
III. RESULTS AND DISCUSSION
II. METHODOLOGY
A. Machine Description and Principle of Operation
A. Design and Fabrication of the Machine The machine was composed of five major components
The machine was designed to perform cleaning by namely: hopper, cleaning unit, grading unit, frame, and transmission
removing excess dry scales and grade red onion bulbs into small (15 assembly powered by a 0.5hp electric motor. The loading of the
to 30 mm), medium (31 to 50 mm) and large (51 to 80 mm) [7]. It uncleaned and ungraded onion bulbs takes place in the hopper. The
was composed of five major components namely: hopper, cleaning cleaning unit, composed of six rollers, was responsible in removing
assembly, grading assembly, frame, and transmission assembly. excess dry-scale and remained one to two dry-scales. Each rollers
Materials was selected based on its cost and availability in the local rotates in opposite direction using spur gears. After cleaning, onions
market. Each sizes was designed based on the allowable stress and are conveyed to the roller-type grading assembly for size grading.
considering the factor of safety of the operation to ensure no failure
156
The total cost of fabricating the machine was ₱125,
000.00. The unit cost of operating the machine was ₱196.22/hr or
₱0.40/kg. As shown in Fig. 2, the machine needs to clean and grade
33,262.16kg/yr to break-even all the costs in 1 year at a custom rate
of ₱1.0/kg in order to consider ownership otherwise it will be
expensive of use. Furthermore, the use of the machine requires 3
persons as operators which lessen the utilization of manual labor
from 0.03 man-day/kg to 0.0008 man-machine-day/kg.
IV. CONCLUSION
Fig. 1. Dry-scale cleaner cum grader for stored onions Based on the results, it can be concluded that:
1. a dry-scale cleaner cum grader for stored red onion bulbs
B. Performance of the Machine was successfully developed based on the considerations
and was fabricated using locally available materials;
Statistically, the machine performance was affected 2. the machine showed satisfactory performance in cleaning
significantly by speed in terms of cleaning capacity, grading capacity, grading capacity, overall capacity, cleaning
capacity, overall capacity, and power consumption at 1% level of efficiency, grading efficiency, overall efficiency, percent
confidence. Results shows that in terms of cleaning capacity, mechanical damage, power consumption, and noise level
grading capacity, and overall capacity, speed in T3 was significantly when set up to 71rpm in cleaning and 17rpm in grading;
higher among T1 and T2. While T1 and T2 were not significantly and,
different. On the other hand, power consumption reveals that T3 was 3. ownership of the machine should be by group of farmers
significantly lower than T1 and T2. The mean power consumption and an individual ownership can also be an option,
when using speed combination of T1 and T2 were not significantly however, it should be subjected for custom service.
different. Similarly, the means of the percent mechanical damage as
influence by different speed variations showed significant difference ACKNOWLEDGEMENT
at 5% level of confidence. Results shows that speed combination of
T1 was significantly lower compared to T3. On the other hand, The researcher wants to express his sincere gratitude to his
means between T1 and T2 as well as between T2 and T3 were not advisory committee Emmanuel V. Sicat Ph.D., Helen F. Gavino
significantly different. Furthermore, in terms of cleaning efficiency, Ph.D., and Theody B. Sayco Ph.D. for their support and guidance
grading efficiency, overall efficiency, and noise level, no significant that leads to the completion of this study. Also, to the Department of
difference was observed in the speed variations on the performance Science and Technology-Engineering Research and Development
of the machine. Results revealed that satisfactory machine for Technology for the scholarship granted to the researcher.
performance was expected when the speed combination were 71rpm
and 17rpm.
157
Semiconductor,
Materials and
Electronics
Session Chair : Dr. Jonathan W.L. Salvacion
Moderator : Engr. Ericson D. Dimaunahan
Panelists : Dr. Michael C. Pacis
Dr. Jennifer C. Dela Cruz
Engr. Febus Reidj G. Cruz
i-LAB-u: Internet-connected Laboratory Units with
Over-The-Air (OTA) Programming Capability
Alexander C. Abad
ECE Department, John Gokongwei Jr. College of Engineering,
De La Salle University, Manila
alexander.abad@dlsu.edu.ph
Abstract—This 2-page abstract provides a step-by-step More detatils about the setup will be discussed in the following
procedures on how to set-up a teleoperated electronics laboratory sections.
with Over-The-Air (OTA) programming capability using
Raspberry Pi Zero with camera, and Particle devices such as
Photon and Argon. This study might help in the online learning B. Hardware
scheme as we battle against the covid19 pandemic by providing The hardware component is based on Particle devices. I
students a hands-on experience on programming remote used Photon and Argon because they have WiFi capability and can
microcontrollers to control different devices such as servos and be programmed anywhere through the use of Web Integrated
LEDs. This will motivate students in learning the basics of OTA Development Environment (IDE) [5]. To demonstrate some output,
programming and Internet-of-Things (IoT). This study can be 2-degrees of freedom (dof) mini-robot made from 2 mini servos
applied to telepresence, remote sensing and monitoring as well as (SG90) is connected directly to Photon1. A tutorial with codes on
in telerobotics applications. how to control a servo using Particle Photon can be found in these
links [6,7]. Moreover, to demonstrate that we can control more
Keywords— Over-The-Air (OTA) programming, Particle than 2 servos, I used PCA9685 (16-Channel 12-bit PWM/Servo
devices, Raspberry Pi, Internet-of-Things (IoT) Driver - I2C interface) [8] connected to Photon2 to drive a 4-dof
mini robotic arm made from 4 mini servos and popsicle sticks.
I. INTRODUCTION There are many video tutorials on how to create a 4-dof mini
robotic arm such as this one [9]. Discussion on how to integrate
Due to covid19 pandemic, most schools were forced to Photon device with PCA9685 can be found here [10].
go online learning or distance learning. Unfortunately, some To demonstrate visual output, I used 8x8 SPI LED
laboratory experiments need actual demonstration or hands-on MATRIX with MAX7219 module to Particle Argon device and
experience. This study aims to help mitigate the hardships in online uploaded the code from this link [11]. The text messages scrolling
teaching and online learning by providing step-by-step process on can be changed through OTA programming using Web IDE [5].
how to set up a teleoperated electronics laboratory with over-the- Aside from servos and LEDs, sensors such as temperature and
air programming capability. The setup in this paper is known as i- humidity sensor can be connected to Particle devices and their
LAB-u for internet-connected LABoratory units. This setup can be readings can be read remotely like in this link [12]. In order to have
used not only for online classes during covid19 pandemic but can a real-time video feedback to see what is happening on the remote
also be as a teaching and learning platform for Over-The-Air setup, I used Raspberry Pi Zero with camera. The total cost of my
(OTA) programming and Internet-of-Things (IoT) in the new setup is around P7,000.00. All the devices can be purchased locally
normal. The setup can be teleoperated anywhere in the world. or online. A rough estimate cost of my setup in PHP:
Particle Photon = P 1,200 per unit
II. METHODOLOGY Particle Argon = P 2,000 per unit
Raspberry Pi Zero = P 850 per unit
A. The Setup Raspberry Zero Pi Mini Camera = P 700 per unit
Matrix LED 8x8 Display module MAX7219 = P 150
The top view and front view of the teleoperated Tower Pro SG90 9g Gear Steering Servo (6PCS) = P 350
electronics lab is shown in Fig. 1. It uses Particle devices Photon PCA9685 = P 200
[1] and Argon [2] with an online camera feedback from Raspberry Breadboard and jumper wires = P 100
Pi Zero [3] with Raspberry Zero Pi Mini Camera [4].
C. Software
My setup uses UNIX for the Raspberry Pi, C++, HTML,
and CSS for the Particle Devices. This paper provides links to
many tutorials to help the reader replicate the setup.
There are so many step-by-step tutorials on how to set up
a Raspberry Pi Zero such as the video tutorial in this link [13] and
this link [14]. After you set up the Raspberry Pi Zero with a
camera, you can make the video feed appear in a browser by
following the instructions in this link [15]. You need to port
forward your local IP network so that you can access or share to
others your camera feed anywhere. Aside from port forwarding,
you can use tunneling through ngrok [16] by signing up a free
Fig. 1. i-LAB-u setup. account. I used ngrok to stream the video of my remote setup.
159
The Particle devices can be programmed like Arduino teaching and learning platform for Over-The-Air (OTA)
using C++ language. Through the use of Web IDE, we can programming and Internet-of-Things (IoT) for the new normal.
program Particle Photon and Argon anywhere. Using HTML and This study can be applied to telepresence, remote sensing and
CSS files like in this link [17], we can create remote control button monitoring as well as in telerobotics applications.
in integrate in a webpage.
Step-by-step procedures on how to setup Particle Photon ACKNOWLEDGEMENT
and Argon can be found here [18] and here [19] respectively.
I would like to thank DOST-ERDT for helping me in my studies.
III. RESULTS AND DISCUSSION
REFERENCES
[1] PARTICLE PHOTON: WI-FI, https://docs.particle.io/photon/
If the tutorial in [14] has been successfully executed, the
(accessed on October 22, 2020)
live video feed from the Raspberry Pi Zero cam can be viewed on
[2] PARTICLE ARGON: WI-FI + BLUETOOTH,
your internet browser connected to your local network as shown in
https://docs.particle.io/argon/ (accessed on October 22, 2020)
Fig. 2. Using ngrok tunneling, you can get public IP and the live
[3] Raspberry Pi Zero,
video feed can be access beyond your local network.
https://www.raspberrypi.org/products/raspberry-pi-
zero/?resellerType=home (accessed on October 22, 2020)
[4] Raspberry Zero Pi Mini Camera,
https://www.lazada.com.ph/products/raspberry-zero-pi-mini-
camera-i207843638-
s263818843.html?spm=a2o4l.searchlist.list.4.2c642c1djDYR
gB&search=1 (accessed on October 22, 2020)
[5] FLASH APPS WITH THE PARTICLE WEB IDE,
https://docs.particle.io/tutorials/developer-tools/build/
(accessed on October 22, 2020)
[6] MAKER KIT TUTORIALS,
Fig. 2. Live video feed from Raspberry Pi Zero Camera. https://docs.particle.io/tutorials/hardware-projects/maker-kit/
(accessed on October 22, 2020)
If the tunneling using ngrok is successful, you will get a link [7] Tutorial: Spark Variable and Function on One Web Page,
similar to this one: http://0.tcp.ngrok.io:13175 that you can use or https://community.particle.io/t/tutorial-spark-variable-and-
share to others for them to see your setup. The tunnel address function-on-one-web-page/4181 (accessed on October 22,
changes whenever you close the Raspberry Pi terminal or when 2020)
you restart it. The live video feed can be embedded inside an [8] Adafruit PCA9685 16-Channel Servo Driver,
HTML by putting these lines inside the body of your HTML code: https://learn.adafruit.com/16-channel-pwm-servo-driver
<center><iframe width="550" height="365" (accessed on October 22, 2020)
src="http://0.tcp.ngrok.io:13175” [9] How to make Micro Servo Robotic arm arduino based simple
</iframe> DIY, https://www.youtube.com/watch?v=MCYJAyIXBdA
(accessed on October 22, 2020)
The live video feed can be combined to the control buttons as [10] Photon + PCA9685 Library,
shown in Fig. 3. https://community.particle.io/t/photon-pca9685-library/22794
(accessed on October 22, 2020)
[11] LedControl-MAX7219-
MAX7221/firmware/examples/matrixpaneldemo.ino,
https://github.com/chrisloris/LedControl-MAX7219-
MAX7221/blob/master/firmware/examples/matrixpaneldemo.
ino (accessed on October 22, 2020)
[12] IOT Particle Temperature Sensor,
https://www.hackster.io/93944/iot-particle-temperature-
sensor-9820b2 (accessed on October 22, 2020)
[13] How to Setup Raspberry Pi Zero W for Headless
https://www.youtube.com/watch?v=LlCr09B2HZI (accessed
on October 22, 2020)
[14] RASPBERRY PI ZERO MODEL W TUTORIAL 2:
MAKING A WIRELESS PORTABLE IP CAMERA
https://toptechboy.com/raspberry-pi-zero-model-w-tutorial-2-
making-a-wireless-portable-ip-camera/ (accessed on October
Fig. 3. Live video feed embedded in a webpage together with 22, 2020)
control buttons. [15] RPi-Cam-Web-Interface https://elinux.org/RPi-Cam-Web-
Interface (accessed on October 22, 2020)
[16] ngrok. https://ngrok.com/ (accessed on October 22, 2020)
III. CONCLUSION [17] Light a IOT LED from web with Particle Photon
https://knsakib.com/2015/12/28/light-a-led-over-the-net-with-
This paper presented step-by-step procedures including
particle/ (accessed on October 22, 2020)
clickable link references on how to setup a teleoperated electronics
[18] Quick start: Photon, https://docs.particle.io/quickstart/photon/
laboratory using Raspberry Pi and Particle devices with over-the-
(accessed on October 22, 2020)
air programming capability. This setup can be used not only for
[19] Quick start: Argon, https://docs.particle.io/quickstart/argon/
online classes during covid19 pandemic but can also be as a
(accessed on October 22, 2020)
160
Design and Development of a Neural Network-based
Coconut Maturity Detector using Sound Signatures
Nemilyn A. Fadchar1, a *, Jennifer C. Dela Cruz2,b
1
Mapua University
2
Mapua University
a
nemilyn_ece@yahoo.com, bjcdelacruz@mapua.edu.ph
Abstract— Fruit maturity detection is one of the factors systematic procedure to determine the maturity stage. Therefore,
that influence the quality of post-harvest products. One of the known predicting the maturity stage of coconut fruit using sound signatures
practices in fruit maturity detection particularly in coconut is by could impact the post-harvest standards in the agricultural sector.
manual tapping or knocking of the surface of the coconut fruit and Moreover, implementation of machine learning techniques could
listening to its sound. Hence, this practice tends to be subjective and establish models that could be used to develop platforms that will
requires skills that is earned through years of experience. improve the traditional manual procedures of the coconut farmers.
Modernizing the postharvest process using low-cost non-destructive This could also facilitate the industrial sorting system as well as
approaches such as acoustics is a promising frontier to uplift the maintain the quality of the end products.
conventional farming practices. Moreover, implementation of According to [2] acoustic properties has been widely used
machine learning tools could further increase the accuracy of such to assess the quality of agricultural products. It plays an important
process. This study attempts to develop a prototype that uses the role in determining different attributes related to maturity in a non-
sound signatures to classify the maturity level of the young coconut destructive approach. Moreover, the adaptation of this technique in
by using neural network. Specifically, it aims to: (1) extract the the post-harvest industry uplift the quality of products and sustain
acoustic features from the sound signal; (2) train the data set and consumer demands with high quality which is beneficial to the
develop the classification model using neural network; (3) design farmers. Several attempts to describe the association between
and develop the prototype; and (4) test and evaluate the prototype. properties and maturity of some fruits has also been explored in
The major parts of the prototype were composed of the vibration different studies. Acoustic signals have been used to classify
motor, vibration sensor, rpi 3b+ microcontroller, battery and LCD. maturity of different fruits such as coconut [1], [3], watermelon [4],
Result of the neural network model obtained an over-all mango [5], cocoa [6], pomelo [7], melon [8] and tomato [9]. Several
classification accuracy of 91.3 %. This implied that the neural studies also utilized the acoustic response to non-destructively
network model has a high prediction rate to accurately determine the determine the internal quality and firmness of different fruits such
maturity level of the coconut using the sound signatures. Lastly, as apple [10], pear [11], pineapple [12], [13] and kiwi [14]. The
final evaluation results showed that the prototype has a higher initial study on coconut maturity detection in the Philippines was
percentage of prediction accuracy as compared to the manual pioneered by the study of [1] to establish a scientific method of
process. determining the maturity level of a coconut fruit. In this study, the
Keywords— coconut maturity, sound signatures, machine young coconut fruit or “buko” were tapped manually by the farmer
learning, neural network, raspberry pi and the sound were collected using a microphone.
The study of [1] use the knocking trick at the coconut shell
I. INTRODUCTION which is the common method of determining the ripeness. The study
utilized an improvised sound box as shown in Figure 1 to collect
Coconut postharvest industry in the Philippines sounds from tapping. A microphone was installed inside the sound
tremendously improved as new technologies were introduced. box and is directly connected to a tape recorder. The analog sound
Coconut which is scientifically known as cocos nucifera L., is one collected was converted using and analog-to- digital converter
of the most valuable crops in the Philippines because different (ADC) and was later converted into its frequency domain using the
products could be derived from this tree particularly in its fruit. The Fast Fourier Transform (FFT). Results revealed that the pitch of
quality of these coconut-based products was maintained through Malauhog was different from Malakanin and Malakatad which
proper selection of coconut fruit according to its maturity stage. One indicates that using the pitch, the Malauhog can be identified. In
of the essential factors in postharvest industry is maturity detection order to further classify the young coconut, the average power was
and classification. Maturity stage of a fruit or vegetable is said to be also used. Results of the experiments done by [1] indicated that
directly related to the product quality.. A coconut fruit must be Malakatad has the highest average power and a finer extent of main
harvested on its proper maturity stage in order to serve its intended spectral component while Malauhog has obtained the lowest average
use as well as maintain its quality. According to [1], an interview power which is supported by its low pitch. This characteristic was
with some of the coconut farmers from the Philippines there were used to differentiate the Malakatad and Malauhog from Malakanin.
actually three known maturity stage of young coconut namely: Further enhancement were done by the succeeding studies such as
Malauhog, Malakanin and Malakatad and these classifications were determination of physical, acoustical, mechanical and chemical
according to meat thickness. The mucus-like thickness which is thin properties of fresh and young coconut for maturity separation [3],
and has relatively transparent color is called Malauhog while the the automated grading of coconut using audioception [15], coconut
second maturity stage was the cooked rice-like, has thicker and solid maturity classification using fuzzy logic [16] and the
meat which is called Malakanin. The third maturity stage which is microcontroller-based coconut maturity prediction [17]. In the
named to be Malakatad has the thickest coconut meat, leather-like study of [15], the proposed methodology uses the concept of
and have stiff to chew quality. Local farmers identify the maturity audioception to term the usability of a coconut, using the sound
stage of coconut by tapping the fruits of “buko” and listening to its produced. A coconut can be segregated into the good or bad
sound. Hence, this practice introduced bias judgement and requires category. The procedure starts by dehusking the coconuts of
several years of experience or practice. Since, quality is directly standard size and shape, then dropping it in a granite slab from a
associated with maturity it is therefore important to establish height of one foot within a noise free environment that has been set-
161
up. The sound produced on impact were recorded. The Fast Fourier Traditional practice of determining the maturity of several
Transform (FFT) acts as the base for all signal related feature fruits including coconut is usually done manually. Hence, this
extraction. FFT was used to convert a signal in wave form into results in human errors, subjective judgement, productivity issues
spectral information from which most distinguished features can be and other several factors that limits the production outputs. As result
extracted using Fourier domain. Features were extracted and of, a growing demand for automated platforms that replace the
classified using the following supervised learning methods namely: traditional method of determining the maturity of several post-
SMO, Dagging and Naïve Bayes according to their pre-defined harvest commodity specially coconut is a challenge to the post-
class. Results shows that sound signals were able to establish a harvest industry. Modernizing the post-harvest process using low-
model for coconut grading which segregates the good and bad cost non-destructive approaches such as acoustics is a promising
coconut. Likewise, in the experiment conducted by [16], image and frontier to elevate the conventional farming practices as well
sound were used to classify the coconut into three (3) stages namely: incorporating machine learning tools to further enhance the accuracy
Malauhog, Malakanin and Malakatad. of the system.
The two input parameters in the fuzzy interference system The general objective of this study was to design and
were obtained using the image analysis and sound spectral analysis develop a neural network-based coconut maturity detector using
as shown in the conceptual framework in Figure 2. The classification sound signature. Specifically, it aims to: (1) extract the acoustic
system used image analysis and sound spectral analysis prior having features from the sound signal; (2) train the data set and develop the
these parameters connected into a fuzzy inference system. The fuzzy classification model using neural network; (3) design and develop
logic classifier will then process the information as per its the prototype; and (4) test and evaluate the protype.
knowledge base and come up with a decision which corresponds to Threats in the global food security is one of the
a maturity stage. The output of the system is any of the coconut consequences of low standards in maintaining the quality of post-
maturity stage: malauhog, malakanin, and malakatad. Based on the harvest commodities. Since maturity is likely to dictate the quality
simulation results, the fuzzy logic classifier was able to determine of the harvest, it is important that a device that surmount the
the maturity stage based of the coconut fruit based on color and limitations of the traditional manual practice be developed to
sound. Meanwhile, the effects of maturity, fruit surface and hitting contribute in uplifting the farming standards. Likewise, this attempt
point on natural frequency were examined by [3] using the acoustic paves the way in applying non-destructive methods such as
property tester. Each sample was applied with an impact force to the acoustics in determining the maturity level of other post-harvest
sample’s equator and the sound signal was recorded through a fruits. Moreover, this study could reveal the potential of using
microphone mounted a few millimeters above the sample surface machine leaning tools increasing the accuracy of automated
180° around the perimeter from the point of impact. Results revealed systems. Lastly, other protypes for specific could also be developed
that the greatest natural frequency difference between the immature as well as to deploy the predictive models in automated platforms to
and the mature stages was 76 Hz and that between the mature and enhance the accuracy of existing systems.
the over-mature stages was observed to be 126 Hz. The natural This study covers the design and development of a
frequencies observed for impact at a certain point for the immature, prototype for predicting the maturity stage of young coconut. It was
mature, and the over-mature stages were 870, 946, and 1072 Hz, focused on extracting the acoustic vibration, pre-processing the
respectively. The recent study of [17], developed a microcontroller- sound signal to obtain the acoustic frequency using the FFT, and
based device for coconut maturity prediction using the Naïve Bayes training the dataset using neural network to create a classification
method. The device is comprised of a noise sensor which is model in order to determine the maturity stage of young coconuts.
connected to a stethoscope diaphragm for collecting the sound signal The maturity stage of young coconut fruit considered were namely
before it is processed in the microcontroller. A digital filter was used Malauhog, Malakanin, and Malakatad. The basic components that
to reduce the noise and determine the average frequency spread of were used to develop the prototype were the following: (1) vibration
sound using the Fast Fourier Transform (FFT) created by knocking motor and vibration sensor for the sound signal acquisition; (2) the
the coconut using the improvised knocking tool. The device was also raspberry pi 3b+ microcontroller was used in order to perform
equipped with an SD card to record the knocking sound frequency. classification using the neural network model; (3) an LCD
In classifying the coconut maturity, the inputs were three (3) average component to display the reading of the maturity phases read by the
frequency values of knocking sound and the size were fed to the system that is “Malauhog”, “Malakanin”, and “Malakatad”; (4) a
Naïve Bayes algorithm in order to perform the prediction. Results rechargeable battery for design portability. Python programming
revealed that the device was capable of classifying the coconut fruit and tensor flow was used to develop a program for the
into three (3) stages namely: young, fairly mature and mature. The microcontroller. However, there were some limitation in the study
system success rate of 80% was obtained from a total of twenty (20) such other parameters or factors affecting the age of coconuts; the
test sample and twenty-four (24) training data. Previous literatures detection of maturity level of a fully-grown coconut fruit, also
shows that classifying ripeness [6], maturity [7] and [4], firmness termed as Niyog; the identification between usable and rotten
[11] and internal quality [12] uses supervised learning tools in order coconuts; and the effect of storage duration after harvesting the
to develop a classification model or establish the regression model coconut fruit.
based on the acquired data set. The study conducted by [6], uses the
Support Vector Machine (SVM) and K-nearest Neighbor (KNN) II. METHODOLOGY
classifier in training the 392 data samples. Eleven (11) feature This section explains the systematic procedure employed in the
vectors with known class labels were used to classify ripe and unripe design and development of the coconut maturity detector protype.
cocoa pods. Results revealed that SVM has a higher accuracy
Figure 4 shows the major procedures involve in the conduct of the
percentage in predicting the ripeness classification as compared to
study namely: sound signature acquisition, neural network
KNN. Likewise, the research conducted by [4] shows that the
modelling, program development, hardware fabrication and lastly
collected acoustic response from watermelons samples were pre-
processed to extract the acoustic features that were used to develop the implementation which includes the testing and evaluation of the
the a non-destructive approach to develop classification system for prototype.
determining watermelon maturity. Results revealed that the neural
network model has a higher accuracy in classifying the watermelon
ripeness as compared with the expert’s prediction.
162
is done layer-by-layer. Each layer was generated by a neuron
and pass on to the next and so on.
163
Upon turning the device on and pressing the scan button, III. RESULTS AND DISCUSSION
the program starts to read the starts to capture the vibration
signal in the tapping, then it will proceed to pre-processing by A. Sound Signatures
obtaining the frequency information though FFT. Hence, if Based on the results the following acoustic frequency
were obtained upon applying FFT on the sound signal. Figure 11
the signal is unclear it will display “TRY AGAIN” in the shows the difference between the spectral magnitudes of Malauhog,
LCD and the user has to repeat the initial procedure. The Malakanin and Malakatad after several series of transformations. It
frequency information is entered to the network model which reveals that Malauhog has the lowest average power and then
performs the classification. But if the frequency information followed by Malakanin, while Malakatad has the highest average
is not obtained it will display “TRY AGAIN” in the LCD and power. This supports the results of [1], that explains the transition
the user has to repeat the tapping process again Finally, the from solid to hollow sound as the coconut matures from Malauhog
program will output the maturity stage and displayed the to Malakatad.
result in the LCD.
The MATLAB neural network model was translated into a
tensor flow model because there is no straightforward method
that allows a MATLAB script to be incorporated in a phyton
script. Then the python script implements the tensor flow
model and reads the data from the input and perform the
classification to produce an output.
D. Hardware Fabrication
The coconut maturity detector was composed of five (5) (a)
major components namely: vibration motor, vibration sensor,
raspberry pi 3b+ microcontroller, rechargeable battery and
the LCD. The vibration motor was used to serve as the
constant source that initiates vibration to the surface of the
coconut fruit while the vibration sensor was responsible for
detecting the sound signal. The Raspberry Pi 3 model B+ was
used because it is capable to perform the machine learning
function that will use the classification model in order to
determine the maturity stage of the coconut fruit and makes
the device stand-alone. A rechargeable LiPo battery was used (b)
to make the device independent from using the AC power
source during its operation for portability purposes. Finally,
an LCD was used to display the maturity stage of the coconut
fruit.
E. Device Calibration
The coconut maturity detector was calibrated to ensure the
accuracy and reliability of the vibration motor and vibration
sensor used in the prototype. The DT2234C non-contact
high/low speed digital tachometer for small motors was used
to determine the motor speed accuracy as well as the
measurement reliability of the vibration sensor. The effect of (c)
distance between the vibration motor and vibration sensor Figure 5. Spectral magnitude of (a) Malauhog; (b) Malakanin;
(c)Malakatad
was also tested prior to the device construction to determine
the appropriate distance between the motor and the sensor. B. Neural Network Model
F. Implementation A total of 300 samples having six (6) acoustic features
each was trained using the neural network. The datasets were then
A preliminary testing was conducted prior to the final modelled through neural network algorithms on MATLAB R2019b.
evaluation of the prototype. In the preliminary testing, all There were 300x6 dataset used as input data for training and 300x3
parts and function were tested if it is working properly. dataset used as target or output data. Data was randomly divided in
Calibration and troubleshooting were also conducted to ten the following allocation: 70% training dataset, 15% validation
(10) coconut fruits that were randomly selected. Lastly, the dataset and 15% testing dataset. Output class 1 stands for Malauhog,
final evaluation was conducted from fifty (50) coconut fruit class 2 is for Malakanin and class 3 is for Malakatad. During the
training stage the following results were obtained 70 samples were
with different maturity stage. Each coconut was labelled
classified as Malauhog, 69 samples were classified as Malakanin
based on the result obtained from the prototype. To further and 53 samples were classified as Malakatad which were accounted
validate the accuracy of the device each coconut was cracked for the accuracy of percentage of 91.4% as shown in Figure 12.
open to determine if it was correctly classified by the device.
164
Figure 7 shows the actual prototype for the coconut
maturity detector. The dimensions of the device were 15 cm x 9 cm
x 3.5 cm. The enclosure was made of a flexible plastic to reduce the
weight but maintain the durability of the device
165
prediction accuracy: 98% for Malauhog, 95% for Malakanin and Phyton program by using the Tensor Flow. Evaluation results
96% for Malakatad. showed that the prototype has a higher accuracy rating of 96.33% as
TABLE II compared to the manual practice with a rating of 89.67.
Some of the Predictions Done Manual vs. Prototype During This study proves that acoustic feature could be extracted
Final Evaluation from sound signatures to determine the maturity stage of young
Sample Manual Prototype Actual coconut fruit. Furthermore, it also reveals that neural network is a
No. Prediction Predictions Maturity Stage significant tool in developing classification models that could be
1 Malauhog Malakanin Malakanin implemented into automated platforms to improve the efficiency of
5 Malakanin Malauhog Malakatad the system.
10 Malakatad Malakatad Malakatad
15 Malakanin Malakanin Malakanin ACKNOWLEDGEMENT
20 Malakanin Malakanin Malakanin
25 Malakatad Malakanin Malakanin The authors would like to acknowledge the Cavite coconut
30 Malauhog Malakanin Malauhog farmers for providing the samples for the study.
35 Malauhog Malauhog Malauhog
40 Malakatad Malakatad Malakatad REFERENCES
45 Malauhog Malakanin Malakanin [1] M. M. Gatchalian, S. Y. De Leon, and T. Yano,
50 Malakanin Malakanin Malakanin “Measurement of young coconut (Cocos nucifera, L.)
maturity by sound waves,” J. Food Eng., vol. 23, no. 3, pp.
253–276, 1994.
100 [2] M. Aboonajmi, M. Jahangiri, and S. R. Hassan-Beygi, “A
Review on Application of Acoustic Analysis in Quality
Evaluation of Agro-food Products,” J. Food Process.
95 Preserv., vol. 39, no. 6, pp. 3175–3188, Dec. 2015.
[3] M. R. Mitchell et al., “Determination of Physical,
Accuracy Percentage
166
acoustics,” in 2000 ASAE Annual Intenational Meeting,
Technical Papers: Engineering Solutions for a New
Century, 2000, vol. 2, pp. 1229–1244.
[13] S. Boonmung, B. Chomtee, and K. Kanlayasiri,
“Evaluation of artificial neural networks for pineapple
grading,” J. Texture Stud., 2006.
[14] T. Y. Yen and D. J. Yao, “Freshness detection of kiwifruit
by gas sensing array based on surface acoustic wave
technique,” in NEMS 2018 - 13th Annual IEEE
International Conference on Nano/Micro Engineered and
Molecular Systems, 2018, pp. 98–101.
[15] P. Thomas and H. B. Anita, “A novel automated method
for coconut grading based on audioception,” J. Theor. Appl.
Inf. Technol., vol. 95, no. 10, pp. 2265–2273, 2017.
[16] I. M. Javel, A. A. Bandala, R. C. Salvador, R. A. R. Bedruz,
E. P. Dadios, and R. R. P. Vicerra, “Coconut fruit maturity
classification using fuzzy logic,” in 2018 IEEE 10th
International Conference on Humanoid, Nanotechnology,
Information Technology, Communication and Control,
Environment and Management, HNICEM 2018, 2019.
[17] D. Rahmawati, H. Haryanto, and F. Sakariya, “THE
DESIGN OF COCONUT MATURITY PREDICTION
DEVICE WITH ACOUSTIC FREQUENCY
DETECTION USING NAIVE BAYES METHOD
BASED MICROCONTROLLER,” JEEMECS (Journal
Electr. Eng. Mechatron. Comput. Sci., 2019.
[18] J. W. Orillo, J. Dela Cruz, L. Agapito, P. J. Satimbre, and
I. Valenzuela, “Identification of diseases in rice plant
(oryza sativa) using back propagation Artificial Neural
Network,” in 2014 International Conference on Humanoid,
Nanotechnology, Information Technology, Communication
and Control, Environment and Management, HNICEM
2014 - 7th HNICEM 2014 Joint with 6th International
Symposium on Computational Intelligence and Intelligent
In, 2014.
[19] T. M. Amado and J. C. Dela Cruz, “Development of
Machine Learning-based Predictive Models for Air Quality
Monitoring and Characterization,” in IEEE Region 10
Annual International Conference, Proceedings/TENCON,
2019, vol. 2018-October, pp. 668–672.
167
Transfer Learning with Convolutional Neural
Network Towards Fungi Classification in Microscopy
Images
Josh Daniel Lao Ong1, a *, Luis Gabriel Cajucom1,b, Erinn Giannice Abigan1,c, Patricia Angela R. Abu1,d and Ma.
Regina Justina E. Estuar1,e
1
Ateneo de Manila University
a
josh.ong@obf.ateneo.edu, bluis.cajucom@obf.ateneo.edu, cerinn.abigan@obf.ateneo.edu, dpabu@ateneo.edu,
e
restuar@ateneo.edu
Abstract— The Panama disease, also known as the Even before infecting a plant host, Foc can be detected in
Fusarium wilt, is a deadly disease known to affect banana plants all infected soil through the use of microscopy. While Foc itself does
over the world. Caused by a fungal pathogen known as Fusarium not consist of any unique fungal structures that would be a sure
oxysporum f. sp. cubense (Foc), the disease has been a constant indicator, the presence of all known structures could suggest that it
threat to banana producers considering that it cannot be eradicated may be present in a given soil sample. This may then be used as a
once it has infected the soil. A new strain that emerged in 1989 low-cost preliminary method in detecting the disease before
called Tropical Race 4 (TR4) is now threatening the Cavendish proceeding with a more comprehensive detection method.
cultivar, the most popular banana variety being grown today. While microscopy would seem to be a feasible method, it
Furthermore, symptoms of the disease are not visible until late stages does require some training in recognizing fungal structures. The
of the infection. While there are methods that accurately determine viability of manually detecting Foc through a microscope also
the presence of Foc in a soil sample, these are costly and inaccessible heavily depends on the scale of the banana production. Because of
to most banana producers. Thus, we propose the use of this, automating this process with the use of image processing and
convolutional neural networks in the automatic detection of Foc machine learning may be an apt solution. In this study, we propose
TR4 in soil samples with the use of microscopy. This study utilized the use of a convolutional neural network (CNN) in detecting the
a dataset containing microscopy images of various fungal species presence of Foc TR4 in microscopy images.
captured using three distinct microscopy techniques: brightfield,
darkfield, and fluorescent. Transfer learning has shown to make
significant improvements to the performance of the models in II. METHODOLOGY
classifying microscopic fungi. The best performing model was
trained exclusively on brightfield images and has achieved an A. Dataset Overview and Description
accuracy score of 93.11%. The dataset used in this study is composed of images taken
Keywords—convolutional neural networks, transfer with a lab microscope of lab-cultivated soil samples. These samples
learning, image classification, fusarium oxysporum, fungi, contain various species of fungi and are labeled accordingly. This
microscopy study will focus on the detection of Foc TR4. As such, the images
were divided into four distinct labels: Clean, Foc TR4, Foc Non-
TR4, and Other Fungi. The number of images under each of these
I. INTRODUCTION labels is shown in Table 1. While the number of images under each
label is not balanced, the use of transfer learning makes it so that
The Fusarium wilt is a disease that has long plagued balancing the number of images would not be necessary. Clean
banana plantations all over the world. It is caused by a fungal images refer to image captures which do not contain any fungal
pathogen known as Fusarium oxysporum f. sp. cubense (Foc) [1]. structure at all while the other labels refer to images that do contain
Since its discovery in the 20th century, numerous efforts have been fungal structures and are captured from microscope slides that are
conducted to combat the spread of the disease. While previous known to contain their respective fungal species.
variants of the wilt have successfully been mitigated, a new strain
that emerged in 1989 has been a sizable threat until today [2]. This TABLE I. DATASET OVERVIEW
strain, known as tropical race 4 (TR4), is known to affect the
Cavendish cultivar of banana which is the most common variety of
banana being traded and consumed in the world as of writing. Label Microscopy Technique Total
The main challenge in addressing the problems brought
about by the disease is that there is no known method of completely
eradicating it once it has infected the soil [3]. Furthermore, Brightfield Darkfield Fluorescent
symptoms of the disease that would be visible to the naked eye are
typically only observed in the later stages of the infection [4]. As
such, there is great importance in methods that would be able to Clean 771 771 771 2313
detect the presence of the pathogen in an earlier manner.
There are methods that can be used in the accurate
detection of the disease. One such method is to use polymerase chain Foc TR4 771 771 771 2313
reaction. However, these methods require expensive equipment and
materials that most small-scale banana producers would not have
access to [5].
168
details in fungal structures as the fluorescence itself, if sufficiently
Foc Non- 876 876 876 2628
intense, may introduce some noise in the image capture. On the other
TR4
hand, brightfield images seem to present the structures in the clearest
detail.
Other 598 598 598 1794
TABLE II. MODEL ACCURACY
Fungi
Fig. 1. Example of an image captured using three microscopy All Techniques Model 90.89%
techniques (from left to right): Brightfield, Darkfield, Fluorescent.
169
TABLE IV. BRIGHTFIELD MODEL METRICS
170
IV. CONCLUSION ACKNOWLEDGEMENT
The early detection of Fusarium oxysporum f. sp. cubense The authors would like to thank the Department of
Tropical Race 4 is crucial towards the mitigation of damages caused Science and Technology - Science Education Institute (DOST-SEI),
by the disease to banana plantations all over the world. Microscopy the Cloud-based Intelligent Total Analysis System (CITAS) Project,
is one method that can be used for the detection of the pathogen, and and the Ateneo Center for Computing Competency and Research
the use of convolutional neural networks is proposed in the (ACCCRe) for their support in this study.
automation of this process. The microscopy images used in this
study utilized three different microscopy techniques in capturing REFERENCES
images of the samples: brightfield, darkfield, and fluorescent
[1] G. Fourie, E. T. Steenkamp, R. C. Ploetz, T. R. Gordon, and A.
microscopy. The effectiveness of these techniques in detecting the
Viljoen,“Current status of the taxonomic position of fusarium
fungi was assessed through the development of four models trained
oxysporum formae specialis cubense within the fusarium
on different subsets of the dataset. Based on the results obtained in
oxysporum complex,” Infection, Genetics and Evolution, no.
this study, brightfield images are more helpful in classifying fungi
11, p. 533–542, 2011.
in microscopy images among the three techniques used in this
[2] M. Reynolds, “A fungus could wipe out the banana forever,”
dataset. It is worth noting that this is true only in the context of the
https://www.wired.com/story/fungus-could-wipe-out-banana-
images used since these do not contain any dirt or other irrelevant
forever/, August 2019.
artifacts. Transfer learning has shown to make significant
[3] R. C. Ploetz, “Panama disease: A classic and destructive
improvements to the performance of the models in classifying
disease of banana,” Plant Health Progress, 2000.
microscopic fungi. The model trained exclusively on brightfield
[4] M. Dita, M. Barquero, D. Heck, E. S. Mizubuti, and C. P.
images was able to attain an accuracy of 93.11% and was able to
Staver, “Fusarium wilt of banana: Current knowledge on
classify three out of the four labels well. Though the model can be
epidemiology and research needs towards sustainable
further improved, it can still function as a low-cost solution in the
management,” Frontiers in Plant Science, no. 9, 2018.
early detection of Foc given that the model was able to distinguish
[5] M. Fakruddin, K. S. Mannan, A. Chowdhury, R. M.
Foc from other species of Fungi relatively well. The model can also
Mazumdar, M. N. Hossain, S. Islam, and M. A. Chowdhury,
be used as a preliminary test before using comprehensive tests that
“Nucleic acid amplification: Alternative methods of
require costly equipment and materials. Future work could include
polymerase chain reaction,”Journal of Pharmacy Bioallied
more experimentation on CNN architectures used, pre-processing
Sciences, no. 5, p. 245–252, 2013.
techniques, and testing on real-world samples gathered from the
[6] C. Tan, F. Sun, T. Kong, W. Zhang, C. Yang, and C. Liu, “A
field.
survey on deep transfer learning,” ICANN 2018, August 2018.
171
Vision-based approach for checkpoint status analysis
Czeritonnie Gail V. Ya-on1, a *, Jonathan Paul C. Cempron1,b, and Joel P. Ilao1,c
1
De La Salle University, Manila
a
czeritonnie_ya-on@dlsu.edu.ph, bjonathan_cempron@dlsu.edu.ph, c joel.ilao@dlsu.edu.ph
172
an example of a parallel queue serving multiple vehicles in multiple The end of the queue is not typically visible from the
lanes over a period of time. camera especially when the vehicles are too far from the camera
view as shown in Figure 4. We approximate the tail position of the
line as the first detection of the object being tracked as it arrives in
the queue.
C. Lane Identification
Data preparation involves partitioning of the tracklet data
points. Partition was implemented because the trajectory starts to get
inaccurate when a vehicle is not completely visible particularly
when it starts to exit the scene. The corresponding vehicle’s center
where tj is a window of time. becomes incorrect when the vehicle is only partially visible. In this
kind of scenario, the cluster results are affected. So for more robust
We generate the trajectories by analyzing the output file analysis, tracklet data points were partitioned according to three
containing the list of tracklets. Lanes are defined as regions in the regions: upper, middle, and lower. Thresholds for separating the
scene passed by vehicle objects that correspond to a cluster of aforementioned regions were based on manual observation of the
trajectories. Inspection areas are points at the lanes that signify a tracklet points and the scene. Each partition was clustered using K-
processing region. We compute the service rate by counting the Means [7]. We choose the cluster partition that produces the best
number of vehicles that pass by the inspection area for a given time outcome which result distinctly represents the lanes. The selected
t. The Q(i,t) is the mean of all Q(i,tj) and time tj can be arbitrarily cluster is used as a basis for imposing membership on the lane.
set, i.e. per hour, half-day, per half-hour, etc. Figure 5 shows a clustered portion of the tracklet points. Colorbar
Arrival rate [5] is the rate of vehicles joining in Li for time represents the cluster grouping or the lane in the scene. We based
unit t and denoted as q(i,t), as shown in (2): the lane membership on majority polling. If most tracklet points in
a partition belong to a certain cluster, say cluster 1, then the rest of
the tracklet points will be imposed on this cluster.
(2)
173
D. Queue Analysis in Quarantine Checkpoints
We gauged the queue status based on the service rate and
arrival rate values of each lanes. Different checkpoint sites were
used in the test to validate how the proposed method works on
different lanes. In Figure 7, we can see in a) that vehicles moves
towards the camera while in c) the vehicles moves away from the
camera. The camera view in b) covers both movements.
In this case, we will use the definition of upstream and downstream
to describe the roads throughout the analysis. Downstream is the
flow of vehicles moving towards the camera. Upstream, on the other
hand, is a flow when vehicles move away from the camera.
a)
(a)
b)
(c)
174
EDSA Starmall
Lane
Saturation
Q q
Level
a.
Video recorded on April 23, 2020 between 9AM-11AM
Fig. 8. Marikina Infanta Highway
TABLE II. QUEUE ANALYSIS RESULTS FOR
One limitation of the experiment is we did not include the CHECKPOINT 2
scenes where the inspection areas are not visible to the camera,. In
Fig 8, we observed that inspection area is undetectable because the
vehicles are too small in the scene. Megamall
Lane
Saturation
Q q
Level
a.
Video recorded on May 18, 2020 between 9AM-11AM
Similarly, scenes like in Fig 9 where the camera is too high are not TABLE III. QUEUE ANALYSIS RESULTS FOR
included in the analysis because the vehicles are undetectable at this CHECKPOINT 3
state.
Guadix
Lane
III. RESULTS AND DISCUSSION Saturation
Q q
Level
The method introduced in this study is investigated using videos Cluster 0 192 216 Under-saturated
showing a checkpoint operations held at EDSA Starmall, Megamall,
and Guadix. The videos were recorded on April 23 (ECQ) , May
18 (MECQ), and June 01 (GCQ), respectively. We used 5-minute Cluster 1 1800 1800 Optimal
chunks of the input videos between 9 AM to 11 AM. The experiment
covers the calculation of service rate, arrival rate, and queue Cluster 2 492 492 Optimal
saturation for each locations; and analysis of trajectories generated
by the algorithm. The colorbars in the figures represent the cluster a.
Video recorded on June 01, 2020 between 9AM-11AM
grouping or the lane in the scene.
In Table 1, the clusters 1 and 2 are optimal; and cluster 0 is over
saturated. Table 2 shows that all clusters are optimal, while Table 3
A. Queue Analysis shows that cluster 0 is under saturated while clusters 1 and 2 are both
optimal.
Tables 1, 2, and 3 show the per-hour results of estimating the service
rate Q and arrival rate q for the selected case site. We evaluate the
queue saturation of each lane by comparing the values of Q and q, B. Trajectory Analysis
according to our queue saturation levels defined in Section B.
Figure 10 shows the generated trajectories for a EDSA Starmall. All
TABLE I. QUEUE ANALYSIS RESULTS FOR clusters are downstream in this case. The lanes in cluster 1 and 0 are
CHECKPOINT 1 merging because there was a roadblock before the inspection area in
cluster 1. It is expected to cause cluster 0 to process additional
175
trajectories. Looking at Table 1, we expected that Q in cluster 0
should significantly increase due to merging. However, it is
observed that some points are uncounted in cluster 0, especially in
the regions where the data points overlap.
176
REFERENCES
ACKNOWLEDGEMENT
177
The ERDT National Conference 2020 was conducted via the video conferencing platform,
Zoom.
178