Modeling and Experimental Verification of In-House Built Portable Ultrafiltration (PUF) System to Maintain Water Quality

Abstract

At present, over 2.6 billion people live without access to a continuous water supply, and nearly 900 million people do not obtain drinking water from reliable sources. To solve these problems, one of this study’s goals is to come up with a water-supply system that uses a simple, inexpensive, portable ultrafiltration (PUF) unit. To determine the effectiveness of the portable system, water-quality analysis has been carried out to determine if the system produces filtered water from various sources of water, reaching drinking-water standards. A simple model of the system using Darcy’s Law was also obtained to predict permeate flux and transmembrane pressure (TMP). Initially, simulation was performed using nominal values taken from the literature for four (4) parameters, i.e., membrane hydraulic resistance, initial rapid fouling constant, mass transfer coefficient, and foulant bulk concentration. By minimizing the error between the model with these nominal values and experimental values, an improved model with updated parameters was obtained using the Evolutionary Programming (EP) approach. With the updated model, the average error between the model and the experiment was reduced from 32% to 9%. This was further validated with new data taken from the experiment. This improved model with the updated parameter was then used to predict the TMP and compared with the experimental value. Contrasting the optimized model with the existing model indicates that the optimized model predicts membrane performance better, leading to a competent and reliable model for the purification of water using a PUF system built in-house.

1. Introduction

Water is essential for sustainable development as well as a key factor for human survival, socioeconomic development, and energy and food production. As highlighted by the United Nations (UN) since 2010, drinking water is one of the major global fundamental human rights [1]. To ensure sufficient, safe, and affordable water access while improving worldwide health, education, and economic productivity, these rights represent major milestones for every nation, including Malaysia. Hence, there is a need to have a balance between human and commercial needs in dealing with water resources, especially considering the high growth of the population worldwide.

Many people in the world consume raw surface and groundwater, hence being likely to become infected with water-borne diseases via contamination with microbial organisms in human and animal waste. Microbiologically contaminated drinking water can transmit diseases such as diarrhea, cholera, dysentery, typhoid, and polio and is estimated to cause approximately 505,000 diarrheal deaths each year, with 73,484 diarrheal deaths in Southeast Asia in 2016 [2] and 261 diarrheal deaths recorded in Malaysia in 2020 [3]. Based on the United Nations Sustainable Development Goals (SDG) on drinking water, everyone should have equitable and universal access to safe and affordable drinking water by 2030 [4]. Therefore, comprehensive water-quality assessment is essential for providing users with the direction to deal with this problem. Geographically, socially, and culturally, there are significant differences between rural and urban areas; those who live in low-income or unofficial settlements typically have less access to good supplies of drinking water than others.

Owing to rapid population growth, urbanization, and rising water needs from agriculture, industry, and the energy sectors, water demand continues to rise. River water quality is deteriorating in urban and rural areas as a result of natural and anthropogenic factors. To manage water quality in river basins, it is crucial to understand the changes and factors affecting river water quality. Natural phenomena, including rock weathering, evapotranspiration, atmospheric deposition, climate change, and natural disasters, all affect the quality of river water. Industrial effluent, household wastes, agricultural practices, including applying pesticides, fertilizers, and manures, and animal husbandry, irrigation, deforestation, and aquaculture, can all be considered anthropogenic causes. In general, almost all sources of water for domestic and human usage must be treated using proper technologies before being made available to the general public since these polluting factors are the primary causes of the reduction in water quality. Water-treatment systems can be divided into two primary categories: conventional and non-conventional. A combination of physical, biological, and chemical processes is used in conventional water treatment, whereas more advanced technology is used in non-conventional water treatment.

However, the consideration of factors related to geography, quality of the water sources, issues with costs, and labor limits the widespread implementation of decentralized and localized drinking-water treatment plants. As of 2020, according to UNICEF, there are 2 billion people, or one in four people, lacking safely managed drinking-water services in the world due to these factors [5]. In Malaysia, about 6% population still do not have access to safe drinking water due to accessibility and economic issues in supplying piped water to these areas [4]. We can reduce this number by supplying our portable system to these areas in Malaysia. Hence, in general, the challenge is to supply an easy-to-transport yet inexpensive, clean water-treatment system by applying appropriate technology to those remote areas lacking a centralized water-supply network. In this research, we demonstrate in detail the performance of an in-house-built portable ultrafiltration (PUF) system, showing that it can produce drinking-water quality from various sources of water. We also perform a simple model of the system using fundamental principles to predict permeate flux and transmembrane pressure (TMP). The model was improved by obtaining optimal parameters of the system using the Evolutionary Program (EP) method based on the error between the model and experimental results.

The scientific novelty of the work is as follows:

(i)

Development of an inexpensive, in-house-built portable ultrafiltration (PUF) system that has been thoroughly evaluated for its performance to see how well it treats different kinds of water while meeting drinking-water quality standards.

(ii)

Formulation of a simple but accurate model for this in-house-built portable ultrafiltration unit, allowing for the prediction of TMP and permeate flux within the portable system. This will allow quick predictions to be made of the performance of the unit without having to run experimental tests all the time.

(iii)

Application of the Evolutionary Programming (EP) approach, to update the parameter of the model to produce an improved model that closely resembles the actual PUF unit and which has been validated by experimental testing.

Section 1 gives an introduction to the problem statement. Section 2 presents a literature review of portable systems that are available on the market. Section 3 details the methodology of the work, and Section 4 discusses the results of the work. Section 5 concludes the paper.

2. Literature Review

Earth’s surface water is mainly found in the ocean (97.25%) and in the polar caps or sea ice cover and glaciers, which make up for 2.05% of total surface water, while the remainder is found in freshwater lakes, rivers, and water from ground resources. There is sufficient fresh water in the world, including water that contains a small quantity of salt (with each liter of fresh water, there is less than 3 g of salt) to meet human needs. However, this freshwater is not available all the time and at all locations as required, and it is also not well distributed globally. Hence, it is important to identify water sources that will help to determine and classify contaminants for water filtration. Semi-permeable membranes are normally used in the process of water microfiltration, ultrafiltration, and reverse osmosis. Microfiltration has pore diameters ranging from 1 micron to 0.1 micron, which can completely block bacteria and partially block viruses. As for ultrafiltration, this type of filter has pore diameters that range from 0.1 microns to 0.005 microns and is capable of completely filtering germs and viruses. For nanofiltration, its purification range is from 0.5 nanometers to 5 nanometers. As a result, nanofiltration is not a suitable choice to desalinate or remove salt from seawater. An Ultra Low-Pressure Reverse Osmosis (ULPRO) membrane has been used to treat brackish water but is not able to treat pure seawater [6].

However, ultrafiltration (UF) is a highly used method in potable water production as this type of filtration filters the total suspended solids (TSS), turbidity, organic matter, microorganisms, etc., from source water [7]. The UF system’s benefits include its capacity to reliably produce high-quality filtrate, its reduced acreage requirements, its non-toxic sludge discharge, and a highly automated procedure that requires fewer workers. The UF system offers promising sustainability since it produces high-quality filtrate without the need for a coagulant [8]. Ultrafiltration uses pressure or a concentrated gradient in order to separate two fluids using a semi-permeable membrane. Since the early 1970s, microfiltration (MF) and ultrafiltration (UF) have become mature separation technologies. The first applications using UF were mainly in the dairy industry, as mentioned by F.A. Glover [9]. Ultrafiltration technology’s advantages over traditional approaches include the UF capability to produce clean water with good quality, operating techniques that are mild, high in selectivity, upgradeable systems that are readily available, and design that is compact and space-efficient [10]. Physical-blockade filtration is used to remove microorganisms from the water, and the hollow-fiber UF membrane technology can provide clean water by using this effective technique. A case study on an industrial-scale drinking-water treatment facility in Malaysia highlighted common design and operational difficulties in assessing the effectiveness of an ultrafiltration (UF) membrane water-treatment plant [11].

Ultrafiltration units can be further classified in terms of their portability and versatility into 3 types, namely, portable, mobile, and modular units. These individual classifications can be seen in

Table 1. The classification of the water-purifying device.

Classification	Description
Portable Unit	Lighter and smaller in size, suitable for single users. It provides drinkable water for individual use.
Mobile Unit	Big in size and a more substantial unit. It is installed on a vehicle and can have a size range from a bicycle to a huge truck or a vessel.
Modular Unit	This unit cannot be transported or moved to new locations without being dismantled and reassembled from the parts at the new locations.

. However, portable purifier units have attracted the most attention from among these classes due to their simplicity of deployment during emergencies, their movability, usage convenience, and ease of maintenance. With only a small amount of investment, portable treatment devices are very useful for applications in households, which provides a feeling of ownership, especially in rural areas. These portable units also have fewer difficulties in their transport and installation [12].

There are, however, only a handful of various types of portable water filters that are available, have varying degrees of effectiveness in the market, and can be utilized with other purification systems.

The Portable First-Response Water Purifier uses a multistage filter that consists of a fabric filter, a sand filter coated with graphene oxide, a vetiver grass filter, and an ultrafiltration system [13]. The Portable Water-Purification Device is a model that combines filter pads that have five layers, which are activated carbon, zeolite, silica sand, mineral sand, and bioball [14]. By utilizing the polyacrylonitrile/biochar (PAN–BC) and polyacrylonitrile/chitosan (PAN–CTN) composite membranes, this device was developed through electrospinning, and subsequently, laccase was then immobilized on the PAN–BC membrane (PAN–BC–LAC).

The Portable Solar-Thermal Purification model is created using polyethyleneimine-grafted-corncob (PEI-g-OC), an agricultural biomass-derived material incorporated in a carbonized carbon dots (CCD) wooden-sponge evaporator. UF membrane facilities in Azuay, Ecuador, were installed by AQUAPOT [15]. These units can meet the drinking-water standards used by the rescue team as well as the disinfected water for medical purposes at a maximum production of 1000 L ${\mathrm{h}}^{-1}$. It uses a hollow-fiber (HF) ultrafiltration (UF) membrane module with a total stand at 100 kDa cut-off.

The Small Portable Water Unit was developed using approximately 500 L daily production capacity. The tubular ceramic membrane, combined with the process of anodic oxidation, was used by the water-treatment mobile unit and is powered by the solar power panel. It is a highly stable production with good results in the sediments, bacteria, colloidal material, and virus removal demonstrated by the results from the tests performed in the laboratory on various kinds of surface water and wastewater treatment plant (WWTP) effluent. The Portable Aqua Unit for Lifesaving (PAUL), known as the “WaterBackpack,” or PAUL, is a compact and lightweight (23 kg) membrane filtering device. It was created at the University of Kassel, Germany, and typically filters more than 1200 liters of water daily. The membrane has a rough lifetime of ten years. Depending on the level of raw water contamination, it is advised to service or clean its filter every few months [16].

There are various commercial portable ultrafiltration devices available on the market at the moment globally and in Malaysia as well. However, details of these devices are limited to their brochures, catalogs, and website, as shown in

Table 2. The commercially available portable ultrafiltration devices on the market.

Product
Name	LG Puri Care	Cuckoo Grande	Panasonic UF Alkaline	Coway Neo	Sawyer	Portawell	Survivor Filter PRO
Process	UF Membrane	Nano Membrane Filter	UF Membrane	RO Membrane Filter	Micro Filtration Membrane	Ceramic Membrane	UF Membrane
Type	4-Stage Filtration	3-Stage Filtration	4 Stages of Filtration	3-Stage Filtration	Tap Filter Type	2 Stages of Filtration	Pump-Typed
Filter Capacity	2 L per Minute	Tank Capacity: 7.6 L	6000 L Capacity	Tank Capacity: 5.8 L	1900 L/Day	230 L per Hour	0.5 L per Minute
Weight	6 kg	18.5 kg	3.8 kg	18 kg	0.15 kg	4.54 kg	0.36 kg
Bacteria Removal	Yes	Yes	Yes	Yes	Yes	Yes	Yes

.

The common features of all the commercially available portable ultrafiltration devices are that they use membrane-based filtration and multistage filtration and are capable of removing bacteria [17,18,19]. The differences noted in each device are the stages, the devices’ filter capacities, and the weight.

Here in our research, we focused on developing a simple, inexpensive, portable UF system that can be tested in the lab on their performance while also developing simple, accurate models that can simulate the actual operation of the system with experimental validation.

3. Methodology

3.1. Description of In-House-Built Portable Ultrafiltration (PUF) Unit

As shown in Figure 1, the filtration system’s equipment is integrated into a transportable unit, which is perfect for medium-scale operation. Pumps, a UF membrane, a UV water sterilizer, and valves are all contained within the box. One pump is used for the normal filtering and the other for backwash, where the UF membrane is utilized for the filtration of the inlet water source and the valve regulates the water flow based on the mode of operation.

The pumps and valves of the portable UF unit are powered by electricity. There are two knobs that we can turn: Knob 1 controls which valve will open or close, and Knob 2 allows us to select the filtration or backwash by activating the appropriate pump. These knobs are situated in the front panel of the device. Others included on the front panels are the digital meters to measure pressure at the outlet backwash, the pressure of the inlet water, the pressure of outlet water, and the flow through the filter. The internals of the PUF can be seen in Figure 2.

3.2. Experimental Setup and Procedure of the Portable PUF System

The experiments were carried out using the membrane module in the dead-end mode, utilizing the system in a single-pass open circuit, as illustrated in Figure 3. Dead-end ultrafiltration (UF) has been regarded as a more energy-efficient method of operation when it comes to producing drinking water with standard quality as opposed to crossflow filtration [20].

Since we are using a backwash procedure, dead-end ultrafiltration mode is preferred over crossflow filtration. For the backwash process, this mode is seen to be superior to crossflow filtration due to several significant operational and performance differences.

The following are the reasons why dead-end filtration is suitable for backwashing:

Solid build-up and cake formation: The entire feed flow is compelled to go through the filter medium in dead-end filtration. Consequently, particles build up and create a “cake” layer on the filter’s surface. Because the particles are stuck on the surface, this cake is effectively dislodged and removed during a backwash, which speeds up the cleaning process. Ease of cleaning during backwash: Backwashing works better to remove the trapped particles because most of the solids build up on the filter surface. Efficiency and throughput during backwash: when the filter surface becomes blocked, the filtration process in dead-end filtration ends, indicating that the system is set up for full filtration. Backwashing is a rapid and effective procedure that frequently restores a sizable portion of the filter’s throughput capacity in a brief period of time.

The feed tank, the pump unit, the UV sterilizer, and the ultrafiltration membrane system are the primary components of this system, which produces drinkable water. All the general parameters of the PUF system, such as the material of the shell, are explained in

Table 3. UF system parameters and operational conditions.

Item	Description
Material of Shell	304 Stainless Steel (food-Grade)
Intake Pressure	1 to 3 Bar
Intake Temperature	5 to 45 °C
Filtration Precision	0.01 Micron
Inlet/Outlet Size (Inch)	0.5 Inches
Backwash Mode	Manual
Membrane Service Life	3 Years (Depending on Water Quality)
Filtration Technology	Ultrafiltration (UF)

.

Before the experiment begins, a test is performed using the raw feed water on the portable system to test the electrical connection and to make sure all equipment, i.e., pump and PUF filter, are functioning well. In the experiment, the water test samples were placed in the storage tank and then passed through the UF unit using the pump located in the box. The filtered water was accumulated and stored in the final storage tank, and this was carried out continuously throughout the experiment. The quality of the water obtained was determined by taking samples from the effluent storage tank and analyzing its turbidity using a turbidity meter, color and TSS using the colorimeter, BOD and DO using the DO meter, pH using the pH meter, COD and ammonia using the photometer, and bacteria using the petrifilm method.

The membrane service life over time, as given by the membrane supplier, is about 3 years, depending on the water quality [21]. Membrane durability can be greatly increased by periodic cleaning with backwashing, which ensures membranes operate under optimal conditions, as shown in our work and described in Section 3.2.

Membrane durability can be enhanced by further treatment of the membranes [22,23], but this will increase the cost of the system and complicate the backwashing process.

Some tests that determine the backwashing duration and frequency have been performed. During the filtration process, there were fluctuations in the flow rate (±15%) compared with the average flow rate, which was caused by the foulants manipulating orientation during the initial step of membrane fouling, when the foulant particles started to narrow down the membrane aperture. The maximum flow rate achieved was 3.95 LPM, when the filtration process was run for about 30 min. After achieving its maximum flow rate, the accumulation of foulants continued, which caused the increase in flow resistance and flow rate to drop gradually until it dropped to 2.36 LPM (a 40% drop from the maximum flow rate), when the filtration process ran for about 65 min. This indicates that the cake formation of foulants starts after about 65 min of the filtration process, which seriously deteriorates the filtration performance. Hence, the backwash system was activated at about the 65th minute to remove the foulants from the membrane. After 4 min of backwashing, the filtration system was reactivated, and the flow rate increased back with a flow rate recovery of 97%. Therefore, for this system, the backwash interval (BWI) for the influent of 25 NTU is every 65 min with 4 min of backwash duration.

As such, the setup of the portable device is ensured, with the connection of the power supply, and all the connection pipes are secured and confirmed with nil leakage. The pump is initiated once the “C” button on the control panel is pressed, after which the sample water is pumped into the unit. The process will then cease once the “OFF” button on the control panel is pressed. By pressing the “B” button on the control panel, the backwashing of the ultrafiltration membrane is accomplished. The system was totally drained and backwashed with clean water for several minutes following each run. For the filtration experiments to be considered reasonably reproducible, each one was run at least twice.

3.3. Water Quality Test

Using a variety of water test kits, we conducted our experiments to check the quality of the water based on major parameters. Subsequently, the WQI is computed to ascertain the treated water’s category concerning its appropriateness for human ingestion. The purpose of the water-quality assessment is to ascertain whether the filtered water (effluent) produced by the portable filter unit satisfies the standards and requirements for drinking-water guidelines established by the WHO and is safe to consume. The quality test and calculation done for the influent and effluent water are given in the next few sections.

The water sample to perform the tests is taken from the Varsity Lake (Tasik Varsity) and river water (Sungai Pantai) located at the University of Malaya. The turbidity of the lake water and river water are measured to be at 16.5 NTU and 24.4 NTU, respectively. However, due to the difficulties in obtaining enough lake and river water to run the flux and TMP results continuously, we have prepared synthetic water as well. The synthetic water was prepared by mixing soil from the Varsity Lake water and manipulating the soil in 60 L of tap water to obtain various turbidity values.

Table 4. Preparation of synthetic water in the lab.

Soil Weight in 60 L Water (g)	Concentration (g/L)	Turbidity (NTU)
10	0.167	17.22
15	0.250	25.30
20	0.333	35.60
30	0.500	51.30

shows the relationship between the soil concentration and the turbidity of synthetic water obtained. The turbidity of the prepared synthetic water turbidity increases proportionately as the soil concentration increases.

We have chosen the synthetic water with a turbidity of about 25 NTU in all future experiments since it is close to the turbidity of the river water, i.e., Sungai Pantai, which is 24.4 NTU. During the experiment for the determination of flux and TMP, this is the water used to top up the river water when it is finished during the continuous experimental run.

Some work on the removal of ammonia nitrogen, a constituent of pesticides, has been done, which shows the capability of our system to reduce its content, as can be seen in [24].

The three types of fouling—organic, inorganic, and biofouling—are caused by pesticides and heavy metals. Pesticides contribute to organic fouling and biofouling, whereas heavy metals mostly cause inorganic fouling (via particle deposition and scaling). In ultrafiltration, fouling is a significant performance-limiting factor that impacts flux decline and increased operational costs.

Detailed analysis of the impact of different levels of contaminants on UF-based system performance has been studied by other researchers for toxic metals, where advanced methods for extracting metals from the contaminated water have been analyzed [25]. In short, extra pretreatment units may need to be added to our system to cater to different levels of contaminants involving heavy metals and pesticides, but with the added cost of production.

3.4. Modeling and Simulation

In this study, the PUF model involves the 800 L per hour membrane capacity applicable from 2 to 3 bars as per specification. The PUF membrane model is made of hollow-fiber UF membrane with a filtration precision of 0.01-micron, intake temperature of 5 to 45 °C and inlet/outlet size of 0.5 inches, as summarized in Table 3, with the principle mode of purification using internal pressure [21].

Concentration polarization and membrane fouling cause the transmembrane pressure (TMP) to rise under constant-flux ultrafiltration. Darcy’s Law can be used to determine the hydraulic reversible resistance and irreversible resistance based on the TMP and flow data [26]. The model below, which is a modified version of the osmotic pressure-resistance model [27], can be used to explain how TMP increases over time:

$$\mathrm{∆}\mathrm{P}=\mathrm{∆}\pi +{\mathrm{J}}_{\mathrm{v}}({\mathrm{R}}_{\mathrm{m}}^0+{\mathrm{R}}_{\mathrm{m}}^{\ast }+\alpha \mathrm{t})$$

(1)

where:

$\mathrm{∆}\mathrm{P}$ = Transmembrane pressure (kPa)

$\mathrm{∆}\pi$ = Osmotic pressure (kPa)

${\mathrm{J}}_{\mathrm{v}}$ = Volumetric permeate flux ($\mathrm{m}/\mathrm{s}$)

${\mathrm{R}}_{\mathrm{m}}^0$ = Membrane hydraulic resistance (kPa s/m)

${\mathrm{R}}_{\mathrm{m}}^{\ast }$ = Initial rapid fouling constant (kPa s/m)

$\alpha$ = Fouling rate constant (kPa/m)

t = Time

Here, ${\mathrm{R}}_{\mathrm{m}}^0$ is dependent on the membrane–foulant system, whereas ${\mathrm{R}}_{\mathrm{m}}^{\ast }$ is a membrane attribute. The present model considers such a process as composed of several short constant-flux phases to explain the variation in permeate flux during constant-pressure ultrafiltration.

This method is justified by the fact that the attenuating character of the permeate flux decline occurs in a process that is under constant pressure, meaning that the rate at which permeate flux decreases is proportional to the decrease in permeate flux magnitude. A previous article [28] used constant-flux studies to experimentally demonstrate that the rate constant for membrane fouling (α) and the osmotic pressure (∆π) were strong factors affecting the permeate flux ${\mathrm{J}}_{\mathrm{v}}$, both increasing with an increase in flux.

The osmotic pressure and fouling rate constants for a specific membrane–foulant system were determined during a series of constant-flux tests. The data from the constant-flux tests is then used to forecast the decline in permeate flux over time for the identical membrane–foulant system operating in the constant-pressure mode. The starting flux in constant-pressure ultrafiltration is assumed in the suggested model to be equal to the flow of pure water (or buffer) of a new membrane at the operating pressure [27].

The system considered the interaction of the three main elements: osmotic pressure (resulting from concentration polarization), resistance caused by membrane fouling, and permeate flux. It takes some time for the foulant’s concentration polarization layer to form. The osmotic pressure model can be used to express the permeate flux:

$${\mathrm{J}}_{\mathrm{v}}={\displaystyle \frac{\mathrm{∆}\mathrm{P}-\mathrm{∆}\pi }{{\mathrm{R}}_{\mathrm{m}}}}$$

(2)

where:

${\mathrm{J}}_{\mathrm{v}}$ = Volumetric permeate flux ($\mathrm{m}/\mathrm{s}$)

$\mathrm{∆}\mathrm{P}$ = Transmembrane pressure (kPa)

${\mathrm{R}}_{\mathrm{m}}$ = Total membrane resistance (kPa s/m)

${\mathrm{R}}_{\mathrm{m}}$ can be expressed as the sum of ${\mathrm{R}}_{\mathrm{m}}^0$ (the resistance of the unfouled membrane) and ${\mathrm{R}}_{\mathrm{f}}$ changes with time due to the deposition and adsorption of foulant. ${\mathrm{R}}_{\mathrm{f}}$ in constant-flux ultrafiltration is expressed as [28]

$${\mathrm{R}}_{\mathrm{f}}={\mathrm{R}}_{\mathrm{m}}^{\ast }+\alpha \mathrm{t}$$

(3)

where:

$\alpha$= m/${\mathrm{J}}_{\mathrm{v}}$

${\mathrm{R}}_{\mathrm{f}}$ = Fouling resistance (kPa s/m)

It was found that ${\mathrm{R}}_{\mathrm{m}}^{\ast }$ was independent of the flux, while m depended on the permeate flux. The current approach is based on transposing Equation (1) to the form shown below, which is proposed for modeling permeate flux changes under constant-pressure ultrafiltration.

$${\mathrm{J}}_{\mathrm{v},\mathrm{i}}={\displaystyle \frac{\mathrm{∆}\mathrm{P}-\mathrm{∆}{\pi }_{\mathrm{i}-1}}{{\mathrm{R}}_{\mathrm{mi}-1}+\left(\left({\mathrm{R}}_{\mathrm{m}}^{\ast }\mathrm{∆}\mathrm{t}\right)/{\mathrm{t}}_{\mathrm{R}}\right)+(\left({\mathrm{m}}_{\mathrm{i}-1}\mathrm{∆}\mathrm{t}\right)/{\mathrm{J}}_{\mathrm{v},\mathrm{i}-1)}}}$$

(4)

where:

${\mathrm{t}}_{\mathrm{R}}$ = Duration of initial rapid fouling phase (s)

∆t = Small time increment (s)

m = Slope of the linear portion of TMP–time profile in constant-flux ultrafiltration (kPa/s)

The cumulative resistance at the time $\left(\mathrm{ti}-1\right)$ is represented as ${\mathrm{R}}_{\mathrm{mi}-1}$. ${\mathrm{R}}_{\mathrm{m}}^{\ast }$ is assumed to have a linear distribution during the time period, which corresponds to the first fast-fouling phase’s duration. This is in line with the conclusion that was mentioned [28]. Here, $\left(\mathrm{mi}-1\right)$ is the slope of the linear region of TMP–time profile in constant-flux ultrafiltration experiment at flux ${\mathrm{J}}_{\mathrm{v},\mathrm{i}-1}.$ While $\mathrm{∆}\mathrm{t}$ is the time step of the ultrafiltration process during which the permeate flux is assumed to be constant.

Equation (4) is used to calculate the permeate flux until t = ${\mathrm{t}}_{\mathrm{R}}$, after which the following equation is used:

$${\mathrm{J}}_{\mathrm{v},\mathrm{i}} ={\displaystyle \frac{\mathrm{∆}\mathrm{P}-\mathrm{∆}{\pi }_{\mathrm{i}-1}}{{\mathrm{R}}_{{\mathrm{m}}_{\mathrm{i}-1}}+(\left({\mathrm{m}}_{\mathrm{i}-1}\mathrm{∆}\mathrm{t}\right)/{\mathrm{J}}_{\mathrm{v},\mathrm{i}-1)}}}$$

(5)

Because of the initial rapid fouling, the resistance rose more quickly in the first few minutes of ultrafiltration. Following this, the rate of rise in fouling resistance decreased over time, making the growth more gradual. This can be understood in terms of the flux-dependent fouling rate, as mentioned by [28], i.e., that the fouling rate falls as permeate flux falls.

Table 5. Model equations.

No.	Equation	Description
1	$\mathrm{∆}\mathrm{P}=\mathrm{∆}\pi +{\mathrm{J}}_{\mathrm{v}}({\mathrm{R}}_{\mathrm{m}}^0+{\mathrm{R}}_{\mathrm{m}}^{\ast }$$+\alpha \mathrm{t})$	To measure the transmembrane pressure (kPa)
2	${\mathrm{J}}_{\mathrm{v}}={\displaystyle \frac{\mathrm{∆}\mathrm{P}-\mathrm{∆}\pi }{{\mathrm{R}}_{\mathrm{m}}}}$	To measure the volumetric permeate flux $(\mathrm{m}/\mathrm{s}$)
3	${\mathrm{R}}_{\mathrm{f}}$$={\mathrm{R}}_{\mathrm{m}}^{\ast }$$+\alpha \mathrm{t}$	To measure the fouling resistance (kPa s/m)
4	${\mathrm{J}}_{\mathrm{v},\mathrm{i}}$$={\displaystyle \frac{\mathrm{∆}\mathrm{P}-\mathrm{∆}{\pi }_{\mathrm{i}-1}}{{\mathrm{R}}_{\mathrm{mi}-1}+\left(\left({\mathrm{R}}_{\mathrm{m}}^{\ast }\mathrm{∆}\mathrm{t}\right)/{\mathrm{t}}_{\mathrm{R}}\right)+(\left({\mathrm{m}}_{\mathrm{i}-1}\mathrm{∆}\mathrm{t}\right)/{\mathrm{J}}_{\mathrm{v},\mathrm{i}-1)}}}$	To measure the permeate flux decline in constant-pressure UF
5	${\mathrm{J}}_{\mathrm{v},\mathrm{i}}$$={\displaystyle \frac{\mathrm{∆}\mathrm{P}-\mathrm{∆}{\pi }_{\mathrm{i}-1}}{{\mathrm{R}}_{{\mathrm{m}}_{\mathrm{i}-1}}+(\left({\mathrm{m}}_{\mathrm{i}-1}\mathrm{∆}\mathrm{t}\right)/{\mathrm{J}}_{\mathrm{v},\mathrm{i}-1)}}}$	To measure the permeate flux decline in constant-pressure UF

summarizes all the equations used to obtain the TMP and the flux of the portable ultrafiltration system.

In this work, the model equations in Table 5 were solved simultaneously to predict the TMP and the permeate flux.

Parameter Estimation Using Evolutionary Programming

The parameters used in these model equations are based on the nominal values as given in the literature [27]. However, the UF system in the literature is slightly different in terms of its properties and characteristics from the portable system in our study. In our study, the parameters concerned are membrane hydraulic resistance, initial rapid fouling constant, mass transfer coefficient, and foulant bulk concentration. Here, the parameters need to be adapted to our system, which is carried out using the Evolutionary Programming approach as described next. Evolutionary computation approaches involve various kinds of algorithms called evolutionary-based algorithms that are inspired by biological evolution in nature [29].

EP based on global path-planning optimization is utilized in this work. The EP with flexibility in the solution representation is an extension of the GA, as suggested by Fogel [30]. In EP, only the evolution process is carried out using mutation operators; there is no crossover operator. Optimal behavior is discovered via robust Evolutionary Programming even during the changing environment. Starting with random strategies, evolution on its own brings about appropriate techniques for solving the current problem [31].

The concepts of natural evolution, particularly the process of natural selection, served as the inspiration for the Evolutionary Programming (EP) optimization process. EP is used to address complicated optimization issues when conventional approaches might not be practical or effective. It is categorized as an evolutionary algorithm, which is a subset of biologically inspired algorithms.

The EP optimization process uses the following key steps:

Initialization: A random population of values is initialized first.

Mutation: Each solution is subjected to mutation in this step, which involves making tiny, arbitrary adjustments to its parameters.

Evaluation: Each member of the population is reassessed to determine its fitness following mutation.

Selection: In EP, choosing the members of the next generation is the next stage. Fitter people are more likely to be chosen because this is determined by their level of fitness.

Reproduction: The newly altered population replaces the old population in this step. By doing this, the public is guaranteed to progress toward better alternatives.

Termination: Until a halting requirement is satisfied or the function’s minimal value is reached, the operation is repeated.

Figure 4 shows the Evolutionary Programming flowchart used to optimize the parameter in our model. First, we do the initialization of the four important parameters, which are Rm0 (membrane hydraulic resistance (kPa s/m)), Rmx (initial rapid fouling constant (kPa s/m)), k (mass transfer coefficient ($\mathrm{m}/\mathrm{s}$)) and Particle 4 is Cb (foulant bulk concentration (kg/${\mathrm{m}}^3$).

The primary variation operator in it is mutation; since individuals inside the population, i.e., the parameter sets in our case, are considered to belong to a certain species rather than to the same species, each parent produces an offspring through a (μ + μ) survivor selection. Producing μ offspring from μ parents is the procedure involved in the (μ + μ) selection strategy. Then, from the combined pool of μ parents and μ children, the best μ set is chosen to produce the next generation of the relevant variables.

There are, in total, five sets of programming to identify the particle. The first set is known as the First Stage. We need to define the parameter as the first step in this stage. The parameters to be optimized are known as particles. We have in total of four particles in this study, as shown in

Table 6. Particles in the First Stage of Evolutionary Programming.

Particles	Description
Particle 1	Rm0 (Membrane Hydraulic Resistance (kPa s/m))
Particle 2	Rmx (Initial Rapid Fouling Constant (kPa s/m))
Particle 3	k (Mass Transfer Coefficient $(\mathrm{m}/\mathrm{s})$)
Particle 4	Cb (Foulant Bulk Concentration $(\mathrm{kg}/{\mathrm{m}}^3)$)

below:

TMP is fixed at 105 kPa, and the time step, deltaT, is fixed at 360. The next step is Initialization, in which five formulas are established. The formulas are to calculate the value for Jv m, α, Cw, and ΔPi. In the subsequent step, we need to start the Initial Program to calculate t(i) and Rm(i) for the i value to be equivalent to 1. In the Main Program step, we are calculating the value for t(j) and Rm(j) to ascertain the value of j from the range of 2 until 25. The second set of programming is known as the Main Body. In this stage, we execute the Initialization, where we define the data value from the ExpData.xlsx file, i.e., the data derived from the experiment. We perform the programming for particles no. 1 until 200, and these 200 particles will go through the programming First Stage. Next, we will calculate the absolute data error for (data-Jv). The sum of errors will be multiplied by 10,000 due to its small value. For each loop, we will clear the $\mathrm{flux} \left({\mathrm{J}}_{\mathrm{v}}\right)$, slope (m), foulant wall concentration Cw, and rate constant x time $\left(\alpha \mathrm{t}\right)$. After this, the third set, known as EP, will be executed.

During this Initialization, the total number of particles is 200, the value of iteration maximum is 100,000 and B is 0.005. For each of the 4 particles, we need to set the problem-specific variables. For Rm0, the problem that needs to be optimized has one parameter, with lower and upper bounds of 100,000 and 500,000, respectively. Concerning the Rmx, the problem that needs to be optimized also has one parameter, with lower and upper bounds of 490,000 and 500,000, respectively. The number of parameters in the issue to be optimized for the k is 1. Its upper and lower limits are set at 0.5 × 10⁻⁵ and 1.0 × 10⁻⁵, respectively. In the case of the Cb, the problem to be optimized also has one parameter, with an upper bound of one and lower bounds of 0.5 for the parameters.

For Set EP, the value xy will be calculated from value 2 until the iteration reaches its maximum. We need to define the particles’ maximum and minimum, the Objective Function’s maximum and minimum, after which we need to mutate Particle 1, Particle 2, Particle 3, and Particle 4 to select the best value, and subsequently mutate the updated particles and perform the Repeated Stage set. Eventually, we will sort and select the best particle, and finally, we will validate if the Objective Function achieved its target. The fifth stage is known as the Repeated Stage, where we repeat the programming of the First Stage but using the new mutated particles. The simulation ends when all four particles (membrane hydraulic resistance, initial rapid fouling constant, mass transfer coefficient, and foulant bulk concentration) converge, i.e., when the sum of error of the response variable is below the set criteria.

4. Results and Discussion

Fouling takes place when there is a build-up of the rejected solute component at the boundary layer close to the membrane surface. Material accumulates on a membrane’s surface (external fouling) or in its pore structure (internal fouling) as a result of fouling in UF. Fouling can result in an irreversible reduction of a membrane’s permeability, lowering the permeate flux and product water quality [32]. The first steps in the fouling mechanism are pore-clogging and pore constriction. This is because the membrane pores contract as a result of the filtered species adhering to them. It encourages the formation of gel or cakes as a result of pore blockage. In most cases, cake formation and pore blockage work together to generate membrane fouling. By preventing the particles from reaching the membrane or by backwashing them out, gel/cake formation can be controlled [33]. In general, adequate backwashing is the most direct way to tackle fouling issues with simple operations, and a detailed study of backwashing on our units has been given in Section 3.2, which will enhance the long-term performance of our system.

In the simulation study, the performance of the membrane unit for the change in transmembrane pressure (TMP) and permeate flux with time is carried out using the models given in Chapter 3. This experiment has been performed in two stages. The first experiment is changing the permeate flux with constant TMP and the second experiment is the change in transmembrane pressure under a constant-flux rate. Once this experiment has been concluded, a comparison using the nominal parameter values versus the improved model with newly updated parameters has been performed in the subsequent steps.

4.1. Improved Modeling Using the EP Approach

Initially, the simulation was performed to obtain the time-dependent response of the flux and TMP through the PUF system based on the models shown in Section 3. This model utilizes parameters based on the nominal values taken from the literature [27,28], which are different from our in-house-built unit in terms of membrane setup and properties. Hence certain parameters contained in the model of the unit must be different as well. The four parameters involved include the membrane hydraulic resistance, initial rapid fouling constant, mass transfer coefficient, and foulant bulk concentration. The results of the model using the nominal parameters can be seen in Figure 5 for the flux rate.

The figure shows the expected flux decrease with time in the PUF due to the membrane fouling as the filtration occurs. However, when the modeling results were compared with the experimental results, the average error obtained was quite high, as seen in Figure 6. This is expected since the model was based on the filtration parameters from the literature which was under different conditions from our own in-house-built system.

The Evolutionary Programming (EP) method was then applied to the difference in error between the modeling and the experimental results to obtain the updated, improved parameters of the model for Particle 1, Particle 2, Particle 3, and Particle 4. See Table 6 for details for each particle. From the optimization cycle using the EP method, the newly updated parameters obtained are shown in

Table 7. Optimized parameters for nominal and improved model.

Parameter Values	Particle 1	Particle 2	Particle 3	Particle 4	TMP
Nominal Values	1.40 × 106	3.86 × 105	9.95 × 10−6	0.3729	105
Improved Values	1.47 × 106	2.68 × 104	9.22 × 10−6	0.0052	105

, which also shows the nominal parameters. The results of the improved model using the updated parameters can be seen in Figure 7, which shows the results for the flux rate to be much closer to the experimental results. Further experimental data were also then taken to validate the improved model to determine the robustness of the improved model. Figure 8 shows the time profiles of flux derived from the improved model and validation data by experiment. For all graphs, every time step represents a 6 min (360 s) time interval. The nominal flow rate of the experimental conditions is at 3.4 L per minute, and the TMP is at 105 kPa.

4.2. TMP Change with Time Using the Improved Model

In this work, the improved mathematical model based on the previous section was performed for the transmembrane pressure during constant-flow ultrafiltration with change in time. The TMP increases with time, which is to be expected due to the accumulation of particles on the membrane i.e., fouling as time progresses (Figure 9). This model to predict TMP was then compared with the experimental value and the results shown in Figure 10, which shows similar profiles and values from both results with an average error of about 9%. The statistical significance of the 9% error reduction is based on the literature reference that the amount of error that is acceptable depends on the experiment, but a margin of error of 10% is generally considered acceptable [34]. The data obtained from the experiment was not smooth since the data are taken at discrete values from the digital instruments to measure TMP and fluctuate slightly, as normally occurs in real experiments.

Hence, in general, it was shown that the prediction of the improved model and the experiment findings agreed well in both permeate flux and TMP predictions in terms of profile and value. This validates the accuracy of the simple model of the PUF system obtained from simple fundamentals and through optimizing the parameters with the EP method.

4.3. Commercial Viability, Sustainability, System Scalability, and Limitations

This system is highly commercially viable due to its low production cost per liter of water produced, i.e., RM 0.10 (USD 0.02) per liter in producing 2000 L of water per day for 5 years. It is designed for portable usage at remote locations at the present moment but can be easily scaled up or down by changing the size of the membrane accordingly with minimum difference in the production cost due to the low membrane cost. This portable system is much cheaper than commercial systems of the same size in the market and other systems quoted in the literature [35], which, however, utilize chemical cleaning, adding to safety issues instead of only backwashing, as in our case. Adequate and optimal backwashing will also enhance the durability of the membranes utilized by other researchers [36,37].

The temperature limit of the membrane in our system is up to 45 °C [21], which is above the normal conditions of water sources in Malaysia. However, if the temperature exceeds this level, pre-cooling may be required prior to entering the system to prevent the polymeric materials of the membrane from deteriorating, but with added cost.

Our system has the option to be driven by solar energy, which addresses the limitation of unstable energy sources.

5. Conclusions

A large population globally, particularly in Malaysia, still does not have access to safe drinking water due to accessibility and economic issues in supplying piped water to remote areas (WHO, 2023). We can reduce this number by supplying portable systems to these areas in a fast, efficient, and economical manner.

Performance tests have been carried out to determine whether the readily available portable ultrafiltration system can filter and treat feed water to fulfill the requirements and guidelines required for drinking water by the Malaysia Ministry of Health (MOH) and the World Health Organization (WHO). Thus, in order to determine the efficiency of the portable unit, water-quality analyses were conducted, and the primary parameters and features that were looked at included turbidity, pH, color, BOD, COD, total coliform, and water-quality index (WQI). The results from this work show that the portable UF device produced drinking water that met DOE and WHO standards. It achieved effluent turbidity below 1 NTU by reducing the turbidity of synthetic water by 99%. Additionally, it achieved effluent color below 15 TCU by reducing the color of synthetic water by 97%. Additionally, it raised the WQI of Class II water sources by 2.4% for synthetic water to become safe Class I drinking water. The portable UF system also showed that it could remove all E. coli and total coliform bacteria from water sources, producing microbiologically safe drinking water. Hence, filtered water from this PUF system is deemed to be safe for human consumption, having met national drinking-water standards.

From the model obtained for the portable UF system, it was found that membrane hydraulic resistance, initial rapid fouling constant, mass transfer coefficient, and foulant bulk concentration were four parameters that had to be optimized using the EP approach when TMP and permeate flux were considered to be response attributes. With the updated model, the average error between the model and the experiment was reduced from 32% to 9%. This was further validated with new data taken from the experiment. This new parameter was also then verified with the model to obtain the TMP. Contrasting the optimized model with the existing model indicates that the optimized model predicts the membrane performance better, therefore making it a competent, fast, and reliable model for the purification of water using the in-house built portable UF (PUF) system. Also, based on the water quality produced by the system and the cost of production, which would be about RM 0.10 (USD 0.02) per liter, in producing 2000 L of water per day for 5 years, it is shown that this system is commercially viable and sustainable for use in remote areas of Malaysia. This is highly acceptable and very much less than many of the other portable commercial systems of the same capacity found in the local market. The system is also highly sustainable since the UV lamp only needs to be changed annually, whereas others need changing as and when necessary. Backwashing is also done for the UV membranes, which makes them usable for a long time, at least 3 years, without changing them.

Future work will include the use of an automatic backwash system to further improve the operation of the unit to make it more practical for rural areas in Malaysia.