Artificial Neural Network Based Modeling of Separation of Xylose from Glucose by Nanofiltration from Monosaccharide Solutions

doi:10.21203/rs.3.rs-3113373/v1

Download PDF

Research Article

Artificial Neural Network Based Modeling of Separation of Xylose from Glucose by Nanofiltration from Monosaccharide Solutions

https://doi.org/10.21203/rs.3.rs-3113373/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

In industrial processes, chromatographic techniques are applied to separate monosaccharides from one another. Here, the investigation was focused on the use of nanofiltration method to separate the pentose sugar from hexose sugar. A xylose is an abundantly available raw material that can be converted into various vital products like xylitol, 2,3-butanediol ethanol and ethanol. The production of xylitol involves xylose as an intermediate product and glucose interferes in this process. The purpose of this research study is to use an Artificial Neural Network (ANN) to create a model for membrane processes based on nanofiltration using already published experimental results. It was interpreted that the xylose and glucose separation using nanofiltration was possible up to a limited extent. It was observed that the mass proportion of a xylose to the glucose was 1.5 to 3.0 times greater in the permeate than their proportion in the feed. Observed monosaccharide retention depends extremely on factors such as permeate flux, type of membrane used for comparable fractionations and the concentration of monosaccharides. The regression model obtained and the corresponding set of simulations strongly support the required results.

Xylose

Glucose

Artificial Neural Network

Nanofiltration

Monosaccharide

Permeate flux

Many studies have been conducted in past on use of different pre-treatment and the hydrolysis methods to convert biomass derived from the oil palm industry into monosaccharides (K. H. Mah H. W., Separation of xylose using a thin-film composite nanofiltration membrane: screening of interfacial polymerization factors, 2016). On an industrial scale, monosaccharide separation is a fairly complicated process. There is a requirement for pure fractions of a particular monosaccharide since many of monosaccharides are key components in pharmaceutical and food sectors (Sjöman, 2007). Due to similarities between the physicochemical properties of monosaccharides, their purification is often carried out via chromatography (K. H. Mah H. W., Optimisation of interfacial polymerization factors in thin-film composite (TFC) polyester nanofiltration (NF) membrane for separation of xylose from glucose, 2019). For instance, the non-charged components xylose and glucose have molecular weight differences of just 150 and 180 g/mol, respectively. Xylose is readily available resource obtained from breaking down (hydrolysis) the hemicellulose component of the agricultural leftovers. It constitutes more than half, or roughly 55%, of total sugar content. Glucose is the another sugar compound that makes up around a quarter, or 25%, of the sugar produced from hemicellulose hydrolysis. Nanofiltration (NF), one of the several separation techniques under investigation right now, offers an affordable and simple-to-maintain alternative for xylose separation from glucose (N F M Roli, 2017). Size of monosaccharide is within size limit of NF membrane. The estimated size of a monosaccharide molecule ranges from 0.6 to 0.8 nanometers. The commercial NF membranes that are commonly used have pore sizes ranging from 0.6 to 2.0 nanometers.Thus, when the total permeate flux changes as a result of changes in pressure or temperature, the relatively tiny monosaccharides are affected mostly (Sjöman, 2007). Some of commercial Nanofiltration (NF) membranes, specifically NF270, Desal-5 DK and Desal-5 DL.have shown a significant capacity to separate monosaccharides, particularly xylose from glucose. The benefits of NF membrane are lower operation pressure and higher permeation flux. Charged and uncharged substances can be separated using nanofiltration membranes.

Some of the factors affecting the nanofiltration separation process include pressure, temperature, composition of monosaccharides and the concentration of total sugars in solution (K. H. Mah H. W., 2014). Higher pressure will result in higher permeate flux, which will lead to a higher rejection of monosaccharide. An increase in pressure causes an increase in solvent flux and membrane compaction, this leads to an increase in the retentions overall, with the greatest impact being on the retention of monosaccharides. Due to decreased viscosity and greater diffusion, retention was observed to decrease with a stepwise temperature increase from 25 to 60 C. The findings also show that a rise in temperature increases the size of polymeric membrane pores and their cut-off size (Sjöman, 2007).

Artificial Neural Network (ANN) modeling of separation of hexose and pentose by nanofiltration was the aim of this research study. Glucose and Xylosewere selected for the hexose and pentose sugars respectively. In order to commercially purify xylose for the manufacturing of xylitol, It is essential to distinguish or isolate the xylose from glucose. From a commercial point of view, most useful and interesting sugar solutions are those that require greater operating pressures due to their higher concentration. Here, total monosaccharide concentrations in the feed solutions ranged from 2–30 wt %, while applied pressure ranged from 2–40 bar. The effects of various solution compositions were investigated while varying the xylose to glucose ratio.

2.1 Experimental Setup:

The experiments were conducted using Frame Filtration Equipment and DSS LabStak M20 plate. The membrane stack consisted of Desal-5 DK, NF270 and Desal-5 DL membranes. All of these membranes are the thin-film composites that can attract and absorb water molecules (hydrophilic).The filtration area of each membrane was 0.18 m². A feed tank is 70 liter reactor which is equipped with steam/water jacket and the mixing arm. A piston pump was used to transfer the liquid to the filtration equipment from the feed tank. A flow meter was used to indicate the rate at which the retentate was returning to the feed tank, while scale was used to measure permeate flux. Membrane NF270 has comparatively low temperature resistance value. Membrane Desal-5 DK carries slight positive electric charge when pH level is 3.5. The pH is close to isoelectric points of Desal-5DL and NF270 membranes (Sjöman, 2007).

2.2 Measurement of pure water flux:

The process began by rinsing a fresh batch of membranes using water that was free of ions, with a conductivity level lower than the 10 µS cm − 1. This was done at the temperature 45°C, pressure 2 bar.
Pressure on membranes was raised slowly to a range of 30 to 40 bar, and they were exposed to a condition of high pressure range for two hours. Afterward, pure water flow rates of untreated membranes were evaluated at 45°C, in water without ions, at pressures of 10 and 20 bar.
Afterward, membranes underwent a cleaning process using an alkaline agent for the 15 minutes at 2 bar and 45°C. Following this, membranes and equipment were treated with water which was ion-free, and once again measurement of pure water flux was taken.
To ensure membranes' cleanliness, Pure water flow rate was evaluated at a temperature of 45°C prior to every filtration process.
To ensure the stability of the membranes used for filtration, a diluted solution containing MgSO4, lactose, and xylose was tested before each filtration of the model solution.This step was taken because the measurement process involved using the membrane of three different sets, with some sets being used multiple times for filtration.. This step was considered important to assess the membranes' ability to retain their filtration properties over time.
In case a membrane showed signs of decline in its ability to separate the materials effectively or if there was a difference in amount of pure water passing through it, it was substituted with a new set of unused membranes.

3.1 Properties of glucose and xylose

Xylose is a raw material coexisting with the other sugars. It can be converted into valuable products like 2, 3-butanediol, ethanol and xylitol by microorganisms. Glucose is a major source of energy for microorganisms.The utilization of a xylose by microorganisms to produce desired result is restricted in the presence of glucose (K. H. Mah H. W., 2014).

A xylose molecule has a molar mass of 150 g mol^− 1 whereas the molecule of glucose has a molar mass of 180 g mol^− 1. Xylose has an equivalent molar diameter of 0.68 nm whereas glucose has a diameter of 0.72 nm.According to the other measurements of size, molecule of glucose is only slightly bigger than molecule of xylose. In water, D-xylose and D-glucose have the same dominant conformation, pyranose ring in the 1e2e3e4e conformation, which means hydroxyl groups connected to the carbons atoms 1–4 are equatorial to a six-membered carbon ring. Both molecules have the ⁴C₁ chair conformation in aqueous solution. The main difference between glucose and xylose in their dominant conformations is that glucose contains "an arm," -CH₂OH. In aqueous solution,62% are as -d-glucopyranose, whereas 38% of the glucose molecules are present as -d-glucopyranose. At 31 C, ratio of - and -anomers is roughly the same as with xylose, with 63% as -d-xylopyranose and 36.5% as -d-xylopyranose. The direction of hydroxyl group at the anomeric center is only difference between - and -anomers. There are two additional structures, namely a furanose ring with 5 members and a open-chain aldehyde. At ambient temperature, In aqueous solutions, there are small quantities of the xylose and glucose present in different arrangements. The proportion of furanose forms in aqueous solution has been observed to increase at a high temperature value. Water molecules have been observed to fit appropriately into existing structure of xylose and glucose. The bonds of hydrogen are formed. Hydration number is a parameter that describes the hydration layer which surrounds a monosaccharide. This refers to quantity of water molecules that are impacted by existence of a monosaccharide (Sjöman, 2007).

3.2 Filtrations of model solutions:

Initially, a mixture was made by combining solutions of pure crystals with ion-free water. The resulting mixture was then dried to form a solid materials that were weighed to be 2%, 10%, and 30% of the total weight. At each level of dry solids, solute of three different type of compositions were created by using xylose and the glucose in mass proportion of 1:9, 1:1, and 9:1. The pH of resulting solution was then lowered to 3.5 by adding concentrated HCl.

The filtration process was carried out under "total reflux mode," where permeate and retentate were sent back or returned to the tank containing the feed. The experiment was conducted on solutions containing 2% and 10% concentrations, and various filtration pressures were tested, including 2, 5, 10, 14, 20, 24, and 30 bar. For solutions with 30% concentration, filtration pressures of 10, 20, 30, and 40 bar were used. The temperature during filtration was kept at 50°C, which was considered safe for both membrane and monosaccharides in solution.

Pressure of feed stream was monitored both before and after it passed through the filtration equipment, and a drop in pressure of 0.5-1 bar was observed due to the presence of the membrane stack. The speed of fluid flow across surface of membrane depended on the thickness of the membrane, with velocities ranging from 0.7-1.0 m/s. To determine the permeate flux,for each membrane, quantity of permeate obtained within a timeframe of 1 to 5 minutes was measured twice at an identical pressure. Before initial measurement of flow rate, a 30-minute period was allowed for system to stabilize at the specified pressure. Total filtration time varied between 1–2 hours for each pressure (Sjöman, 2007).

The measurement of retentate flow was done using a device called flow meter. From permeates, retentate, and feed at each pressure level samples were taken. The xylose and glucose content of samples was analyzed using HPLC technique.

The pH and conductivity of specimens were examined and evaluated. Following each filtration of model solution, membranes were thoroughly washed with water that did not contain ions, and then cleaned using acidic and alkaline cleaning agents. Acetic acid was utilized as acidic cleaning agent, and basic cleaning procedure was conducted in same manner as it was with a batch of new membranes (Sjöman, 2007).

Machine Learning (ML) is a group of computer-based techniques that can identify patterns within extensive collections of data. These templates can be used to learn various nonlinear process, mathematically model them, and predict outcomes. ANNs are one such tool. As name implies, an Artificial Neural Network (ANN) is network of computational processor known as nodes or artificial neurons. These are connected together to form a non-linear processing system. (Alpaydin, 2014)

A multilayer perceptron (MLP) is type of neural network that has multiple layers of neurons. The first layer, called the input layer, receives input data and processes it. The processed data is then sent to one or more hidden layers for further processing. Subsequent output layers perform final output calculations (Sarkar, 2009). The connection between the inputs and the outputs of a neural network is known as its architecture or topology (Martin T. Hagan, 1997). Configuration of an artificial neural network can be altered by modifying quantity of neurons present in hidden layers or by changing number of hidden layers altogether. For this particular study, an ANN configuration consisting of two input neurons and two output neurons was utilized.

Three sets of input data are typically divided by ANN: training, validation, and testing. During the training process, ANNs use the training set to learn patterns and develop their models. Validation set is used to fine-tune the model, ensuring that it can generalize well to new data. Once a model is fully trained, test set is used to evaluate its final performance and compare it to other models (R. Porrazzo, 2013).

In order to obtain a model which furnishes us with the output with least mean-squared error we have used a MATLAB code which provides us with a graph which have number of hidden layers on the x-axis and the value of Mean Squared Error (MSE) on the y-axis. The minima obtained from the graph is the value of the hidden layer which we are going to use to train our model. The corresponding graphs are:

4.1 Data Extraction

Plots, graphs, and diagrams are graphical representations used to describe raw data collected from observations in experiments and scientific publications. xyExtract software was used to extract raw data (from 2D plots) from published research articles. The extracted data is then used for ANN and simulation training to produce findings for reaching optimum concentration based on variations in output and input parameters at different operating circumstances, namely speed and time. The experimental data from (Sofie T. Morthensen, 2015)are used to train the ANN in this work.

4.2 Details of the simulations

The ANN model was built with MATLAB's Deep Learning Toolbox (Kim, 2017). A total of 176 experimental data findings are used for training, validation, and testing the Artificial Neural Network (Sofie T. Morthensen, 2015). Ultimate artificial neural network (ANN) model consists of three layers: an input layer having the same quantity of neurons as that of input parameters, a hidden layer with a variable number of neurons, and an output layer containing only one neuron, which is linked to a single output.

To determine the best topology for an artificial neural network (ANN), several models were trained with varying numbers of neurons in the hidden layer. Model that had the least mean squared error (MSE) was chosen to forecast output parameters. Using MATLAB, anticipated outcomes were graphed against input parameters, resulting in creation of three-dimensional surface plots.

The trained artificial neural network (ANN) is utilized in combination with MATLAB optimization toolbox to determine most efficient values of operating parameters. The ANN is executed in a separate MATLAB script from the optimizer. This script makes predictions of the concentrations for a particular set of input variables, which can be used to calculate objective function's value as output variables. The script is designed as a MATLAB function that is run in the MATLAB Optimization Toolbox to predict the optimal values for operating parameters. To maximize concentration value, simulations are carried out at different concentrations and operating times.

We implemented Sobol's sensitivity analysis in our study. A statistical method called sensitivity analysis that applies Sobol's method allows one to measure the contribution of each input variable (or factor) to the uncertainty and variability of an output from a complex system. The basic goal of Sobol's analysis is to separate the sum of variances provided by each input variable and its interactions from the overall variance of the output. The technique is based on creating inputs that are randomly selected and calculating the model's output for each set of inputs. The following phases make up the approach for sensitivity analysis using the SAlib (Sensitivity Analysis Library) package:

Define the model: We defined the model that we wanted to analyze. This model could be a mathematical equation, a simulation, or any other type of computational model.

Specify the input variables: Next, We specified the input variables that we wanted to include in the sensitivity analysis. We specified the distributions of the input variables, including their lower and upper bounds.

Sampling the inputs: SAlib provides several methods for sampling the inputs, including Latin Hypercube Sampling and Monte Carlo sampling. We created a set of input samples using a way that fulfilled our needs.

Evaluate the model: Once We had the input samples, we evaluated our model for each set of inputs and calculated the corresponding output values.

Calculate the sensitivity indices: SAlib provides several algorithms for calculating the sensitivity indices, including the Sobol indices, the Morris method, and the FAST method. These indices quantify the contribution of each input variable to the variability and uncertainty of the output.

Visualize and interpret the results: The final step was to visualize and interpret the sensitivity analysis results. SAlib provides several visualization tools and methods for understanding the results, such as plots and tables.

To fully comprehend the model's sensitivity to changes in the inputs, this process is repeated for various sampling techniques and sensitivity analysis algorithms. Additionally, SAlib offers several techniques for performing uncertainty analysis and merging the outcomes of many sensitivity tests.

The current research employs an ANN to develop a model for nanofiltration-based membrane processes. The objective is to separate xylose from glucose present in concentrated monosaccharide solutions with a range of 2–30 wt.%, at varying applied pressure (2–40 bar) and temperature of 50◦C. The membrane stack consists of three different membrane - Desal-5 DK, NF270 and Desal-5 DL - with comparable hydrophilic traits. To train and validate the ANN, experimental data from a previous study on hexose-pentose separation are utilized.

Table 1. a) Data extracted to observe a difference between the observed glucose and xylose retention for 2, 10 and 30 wt % feed monosaccharide concentration at total permeate flux below 100 kg m^− 2 h^− 1. Mass ratio of xylose to glucose in feed solutions 9:1. Desal 5-DK(1), Desal 5-DL(2), NF270(3) membranes were used.

(b) Data extracted to observe difference between observed glucose and xylose retention for 2, 10 and 30 wt % feed monosaccharide concentration at total permeate flux below 100 kg m^− 2 h^− 1. Mass ratio of xylose to glucose in feed solutions 1:9. Desal 5-DK(1), Desal 5-DL(2), NF270(3) membranes were used.

Artificial Neural Network Based Modeling of Separation of Xylose from Glucose by Nanofiltration from Monosaccharide Solutions

Status:

Version 1

Abstract

Figures

1. INTRODUCTION

2. THEORY

2.1 Experimental Setup:

2.2 Measurement of pure water flux:

3. MODEL SOLUTIONS: XYLOSE-GLUCOSE

3.1 Properties of glucose and xylose

3.2 Filtrations of model solutions:

4. ARTIFICIAL NEURAL NETWORK

4.1 Data Extraction

4.2 Details of the simulations

5. SENSITIVITY ANALYSIS

6. RESULTS AND DISCUSSION