You are on page 1of 21

EXPERIMENTAL DESIGN FOR ADSORPTION (BOX WILSON DESIGN)

AN ASSIGNMENT SUBMITTED BY NWOKOMA DARLINGTON, B. (REG. NO.: 20074609578)

IN PARTIAL FULFILLMENT FOR THE REQUIREMENT FOR THE COURSE CHE 705: APPLIED STATISTICS IN CHEMICAL ENGINEERING RESEARCH

FEDERAL UNIVERSITY OF TECHNOLOGY, OWERRI (FUTO) IMO STATE, NIGERIA.

1.0 1.1

INTRODUCTION Research Background

The Design of Experiment (DoE) is an efficient procedure of combining two or more variables and point out the correlation amongst them. It involves planning experiments so that the data obtained can be analysed to yield valid and objective conclusion. There are different types of DoE and their choices depend on the purpose of the research (comparative aim, screening or building a model). The most popular of DoE is the Box-Wilson design aka the Central Composite Design (CCD). The Box-Wilson experimental design is a response surface method used for evaluation of a dependent variable as a function of independent variable. In order word, it is an empirical modeling technique dedicated to evaluating the relationship of a set of controlled experimental factors and their observed responses. 1.2 Research Objective and Scope

The aim is to obtain a well-validated quadratic model which would allow prediction of adsorptivity values in an adsorption experiment, as a function of time and sorbent mass, within the working limit of those variables. The objective of this work is to apply BoxWilson experimental design to adsorption experiment data and investigate the response of yield to varied adsorption parameters. In this work, the variation of adsorption capacity as a function of time and mass of adsorbent of an adsorbent that is washed with two different kinds of solvents (5%H2SO4 and Water) is investigated. Thus, it implies studying a real function with two independent variables, preferably using the Box-Wilson design, which is a surface response method. 1.3 General Overview

Chapter 1 is an introduction of the research work. Chapter 2 discusses the theoretical background of the work. The application of Box-Wilson design to analyzing data obtained from adsorption experiment is divulged in Chapter 3, while Chapter 4 highlights and analyses the results obtained from the application of Box-Wilson design. Chapter 5 concludes and makes recommendations based on the findings.

2.0 2.1

THEORITECAL BACKGROUND Design of Experiments: an synopsis

Design of Experiment (DoE) is a structured, systematic and rigorous approach to problem solving that applies principles and techniques at the data collection stage, so as to ensure the generation of valid, defensible and supportable conclusions (Offurum and Chukwu, 2011; Matthews, P. G., 2005). It is a planned approach for determining cause and effect relationships (Liza, Z. R., 2004).It is the laying out of detailed experimental plan in advance of doing the experiment (NIST/SEMATECH, 2010). In chemical engineering industry, it is a method used to determine the empirical relationship between the different factors affecting a process and the output of that process. Establishing such relations enable specification of variables mixture that would achieve some practical benefits. In an experiment, one or more process factors (or variable) are deliberately changed in order to observe the effect the changes have on one or more response variables. The statistical design of experiments is an efficient procedure for planning experiments so that the data obtained can be analyzed to yield valid and objective conclusion. 2.2 Methodology of Design of Experiment

Design of experiments (DoE) deals with experimental methods, thus it has its own terminology, methodology and subject of research. The methodology of DoE was introduced by Fisher in 1935 with six basic principles (Mason et al., 2003;Lazic, R. Z., 2004; Vahid et al., 2011; Fisher, R., 1935) namely; comparison, replication, blocking, randomization, orthogonality and factorial experiments. Comparison is the use of a control which acts as a baseline. Replication is the use of multiple measurements for each experiment which is good mean for the estimation of casual variation around the results. Blocking is the arrangement of experiments into groups (also called blocks). Blocking reduces sources of variation, thus allowing a more precise estimate of the output of the process. Randomization is the process of assigning the various levels of the investigated factors to the experimental units in random order. When a set of experiments is divided into different groups or blocks, randomization gives each experiment the same choice of being assigned to any of the groups. Orthogonality which means that there are several set of experiments or some levels of the factors that are independent of one another and can be combined to derive all the combination of the design. Factorial experiments, i.e. the use of
3

significant combinations amongst the different factors, instead of evaluating one factor at a time. Box et al., (2005) gave a systematic description of the step-by-step procedure for effective DoE approach. Figure 1 illustrates a robust DoE approach for chemical engineering processes.
Select the problem

Determine dependent variable

Determine independent variable

Determine possible combination

Data collection

Modeling

Validation of the Model

Figure 1: Step-by-step procedure for DoE approach in Chemical Processes

Depending on the purpose and the number of factors to be tested, the basic types of experimental design are full factorial design, fractional factorial design, central composition design, which is also known as Box-Wilson, Plackett-Burman design and Box-Behneken design. The purpose of the DoE may be to screen/study the effects of two or more factors on a process and its output or the building of model/optimizing the output of a process. 2.3 Model for Design of Experiment

The purpose of statistically designing an experiment is to collect the maximum amount of relevant information with a minimum expenditure of time and resources. Experimental data are used to derive an empirical (approximation) model linking the outputs and inputs. Such models usually contain first and second order terms. The most common empirical models fit
4

to the experimental data take either a linear form or quadratic form. A linear model with two factors, X1 and X2, can be expressed as

where Y = response for

given levels of the main effects X1 and X2, X1X2 =possible


= cross product coefficient, = experimental error.

interaction effect between X1 and X2, 0 (constant) = response of Y when both main effects are 0,1and 2=linear coefficients,

A quadratic model is a second order model. Such model with two factors, X1 and X2, can be expressed as

where =

= quadratic coefficients.

A second order model (polynomial) is formed during second order designs for describing a response surface. The second order designs are practicable in situations when the linear model is insufficient for a mathematical description of a research subject with adequate precision. We also have higher order polynomials like the cubic or third order polynomials, which is defined by

2.4

RESPONSE SURFACE METHODS

The response surface method basically involves three major steps. The first step is to properly design the experiment in order to evaluate model parameters efficiently after performing experiments. The second step is to develop a polynomial equation to which the experimental data through regression is fitted (i.e. response surface modeling through regression). Then test the correlation fitness by applying statistical criteria and finally evaluating the response by the fitted model (i.e. predicting the response and checking the adequacy of the model). The response surface method enables the representation of independent process parameters in quantitative form as:

where

= predicted response (yield),

= response function,

independent variables, and = experimental error.


5

The surface response is obtained by plotting the expected response of

. The form of

is

unknown and may be very complicated. Thus, a response surface method aims at approximating by a suitable lower-ordered polynomial in some region of the independent

process variables. If the response can be well modeled by a linear function of the independent variables, the function in Equation (3) can be written as (Lazic, 2004; Mason et al., 2003; Montgomery, 2005;Vahid et al., 2011): = ith linear coefficient, = independent variable. = ith quadratic

where k = number of factors of the design, coefficient, = ith interaction coefficient,

The predicted response ( ) is therefore correlated to the set of regression coefficients ( ): the intercept ( ), linear ( ), interaction ( ) and quadratic coefficients ( ).

2.5

BOX-WILSON EXPERIMENTAL DESIGN

The Bow-Wilson design is a response surface method which is an empirical modeling technique devoted to the evaluation of the relationship of a set of controlled experimental factors and observed results (Myers and Montgomery, 2002). This popular second order rotatable experimental design, which is also known as Central Composite Design (CCD),enables experiment to be designed in such a way that factors are varied on three or more levels, rather than on only two level. The Box-Wilson design does not require large number of design points, therefore the cost and time needed for performing experiment is reduced. The Box-Wilson design involves five different levels ( and contains a factorial design central point ) for each factor

, added with some combinations containing the )

(i.e. the mean value of each factor) and some star or axial points(

(i.e. the minimum and the maximum of the range) to allow a good estimation of the curvature of the output. A central composite design always contains twice as many star (or axial) points as there are factors in the design. The axial points are the new extreme values. There are three types of Box-Wilson Design (Circumscribed, Inscribed and Face-Centered), depending on the distance of the levels from the average value of each factor; the most used in chemical engineering is the Circumscribed Design, typically referred to as Central
6

Composite Design (CCD). In a CCD the distance from the centre of the design space to a factorial point is 1 unit for each factor, the distance from the centre of the design space to a star (or axial) point is with | | The precise value of depends on certain properties

desired for the design and on the number of factors involved. If the design is rotatable (i.e. the levels can be put on a circumference and the central point is the centre of this circumference), the value of , can be calculated as; [ 2.5.1 1. 2. ]

Step-by-Step Procedure for developing Box-Wilson design

Define the variables (i.e. the factors) Define the levels of the variables (i.e. choose a range for each factor), which is

usually set/coded as - and + for the minimum and maximum respectively. Then, the mean value (i.e. the central point of the range) coded 0, is considered. Finally, the mean value of the ranges - /0 and 0/+ are estimated (i.e. -1 and +1, referred to as the

levels). For example, if the researcher is studying the combined effect of pH and temperature on microbial degradation of BOD in wastewater treatment plant. Knowing that microbial activity strive in the temperature range of 30oC - 35oC, therefore, 32 and 35 will be referred to as the coded levels - and + , respectively. Then, the mean value of this range (32.5oC) is the central point of the experimental design (level 0). Finally, the researcher calculates the mean values of the ranges 30oC - 32.5oC and 32.5oC - 35oC, that are 31.25oC and 33.5oC; these values are referred to as -1 and +1. 3. Define the number the number of combinations of the design: The number of ) or

experiments in Box-Wilson design contains three sets including; factorial runs ( fractional factorial , studying factors at -1 and +1 level; centre point runs (

examining factors at the centre point of the design space, helping in understanding of curvature and replicating them to evaluate pure errors and axial or star point runs ( setting all factors to 0 (i.e. the centre point) except one, which has the value and + )

(Lizac, 2005).The number of experimental combinations Ncan be calculated as(Lizac, 2005, Montgomery, 2005): (6) where = number of central point runs.

A 2-factor CCD requires at least 11 combinations (normally 13), a 3-factor CCD requires at least 15 (normally 17), whereas a 4-factor design are at least 25 (normally 30). 4. Write the combinations, adding to the design the so called cube or factorial points,

the star/axial point and null/central points. The cube/factorial points are the combinations of the design in which all factors are set to the levels -1 or +1, following a binary procedure; a central point is a kind of control of the CCD, because all the factors are set to the mean value of the range, i.e. the level 0. The star point contain the factor at the minimum or maximum (i.e. the level - and + ) and the other set to the central value. 3.0 3.1 PROBLEMSTATEMENT AND SOLUTION Box-Wilson Experimental Design for Adsorption

The experimental data collection points (Box-Wilson) are shown in Tables 1 5 below. Table 1: EXPERIMENTAL DATA COLLECTION POINTS (BOX WILSON)
LEVEL PROCESS VARIABLES TIME CONC ACTUAL VALUE DESCRIPTION CODE TIME Mass g/l YIELD

1 2 3 4 5

5 13.0546 32.5000 51.9454 60.0000

100 158.5786 300.0000 441.4214 500.0000

Extreme Intermediate Centre Point Intermediate Extreme

-K -1 0 1 K

5.0000 13.0546 32.5000 51.9454 60.0000

0.500 0.793 1.500 2.210 2.500

Table 2: AXIAL POINTS TIME -K K 0 0 CODED FORM CONC 0 0 -K K S/NO 1 2 3 4 ACTUAL VALUES TIME Mass g/l 5.0000 1.5 60.0000 1.5 32.5000 0.5 32.5000 2.5

Table 3: FACTORIAL POINTS CODED FORM TIME MASS g/l -1 0 -1 0 1 -K 1 K ACTUAL VALUES TIME Mass g/l 5.0000 1.5 60.0000 1.5 32.5000 0.5 32.5000 2.5
8

Table 4

3 CENTRE POINTS ACTUAL VALUES TIME Mass g/l 5.0000 1.5 60.0000 1.5 32.5000 0.5 32.5000 2.5

CODED FORM TIME CONC -K 0 K 0 0 -K 0 K

Table 5 S/NO 1 2 3 4 5 6 7 8 9 10 11 12

COLLECTIONS CODED TIME -K K 0 0 -1 -1 1 1 0 0 0 0 MASS g/l 0 0 -K K -1 1 -1 1 0 0 0 0 ACTUAL TIME MASS g/l 5 1.5 60 1.5 32.5 0.5 32.5 2.5 13.1 0.793 13.1 2.21 51.95 0.793 51.95 2.21 32.5 1.5 32.5 1.5 32.5 1.5 32.5 1.5

X1 = 5, 13, 33, 52, 60 X2 = 0.5, 0.793, 1.5, 2.21, 2.5 3.1.1 Adsorption with Adsorbent (ZW1) Tables 6shows the experimental data collection point (Box-Wilson) for adsorption by adsorbent (ZW1) washed with 5%H2SO4, respectively, at temperature, T = 32oC, stirring time X1, mass of adsorbent (g) and initial concentration of 7167.5mg/l. Tables 7 shows the experimental data collection point (Box-Wilson) for adsorption by adsorbent (ZW1) washed with water at temperature, T = 32 oC, stirring time X1, mass of adsorbent (g) and initial concentration of 7167.5mg/l.

Table 6: Experimental Data Collection Points (Box-Wilson) for Adsorbent washed with 5%H2SO4at T = 32oC (ZW1) ACTUAL VALUE Y S/N Time Mass of Time Mass of Conc. (Min) Adsorbent (Min) Adsorbent (mg/l) X1 X2 X1 X2 1 -K 0 5 1.5 1687.95 2 K 0 60 1.5 1218.48 3 0 -K 32.5 0.5 1435.29 4 0 K 32.5 2.5 828.74 5 -1 -1 13.1 0.793 1664.29 6 -1 1 13.1 2.21 1257.83 7 1 -1 51.95 0.793 1338.17 8 1 1 51.95 2.21 633.06 9 0 0 32.5 1.5 1364.51 10 0 0 32.5 1.5 1364.51 11 0 0 32.5 1.5 1364.51 12 0 0 32.5 1.5 1364.51 Table 7: Experimental Data Collection Points (Box-Wilson) for Adsorbent washed with water at T = 32oC (ZW1) CODED VARIABLES S/N ACTUAL VALUE Y Time Mass of Time Mass of Conc. (Min) Adsorbent (Min) Adsorbent (mg/l) X1 X2 X1 X2 -K 0 5 1.5 1759.62 K 0 60 1.5 1318.82 0 -K 32.5 0.5 1416.30 0 K 32.5 2.5 1236.40 -1 -1 13.1 0.793 1650.03 -1 1 13.1 2.21 1598.35 1 -1 51.95 0.793 1382.75 1 1 51.95 2.21 1226.36 0 0 32.5 1.5 1332.22 0 0 32.5 1.5 1332.22 0 0 32.5 1.5 1332.22 0 0 32.5 1.5 1332.22 CODED VARIABLES

1 2 3 4 5 6 7 8 9 10 11 12

10

3.1.2 Adsorption with Adsorbent (ZW2) Tables 8 shows the experimental data collection point (Box-Wilson) for adsorption by adsorbent (ZW1) washed with 5%H2SO4, respectively, at temperature, T = 32oC, stirring time X1, mass of adsorbent (g) and initial concentration of 7167.5mg/l. Tables 9 shows the experimental data collection point (Box-Wilson) for adsorption by adsorbent (ZW1) washed with water at temperature, T = 32 oC, stirring time X1, mass of adsorbent (g) and initial concentration of 7167.5mg/l.
Table 8: Experimental Data Collection Points (Box-Wilson) for Adsorbent washed with 5%H2SO4 at T = 32oC (ZW2) CODED VARIABLES ACTUAL VALUE Y S/N Time Mass of Time Mass of Conc. (Min) Adsorbent (Min) Adsorbent (mg/l) X1 X2 X1 X2 1 -K 0 5 1.5 1698.70 2 K 0 60 1.5 566.23 3 0 -K 32.5 0.5 1644.22 4 0 K 32.5 2.5 151.24 5 -1 -1 13.1 0.793 1812.00 6 -1 1 13.1 2.21 1164.72 7 1 -1 51.95 0.793 1218.48 8 1 1 51.95 2.21 129.02 9 0 0 32.5 1.5 759.04 10 0 0 32.5 1.5 759.04 11 0 0 32.5 1.5 759.04 12 0 0 32.5 1.5 759.04 Table 9: Experimental Data Collection Points (Box-Wilson) for Adsorbent washed with water at T = 32oC (ZW2) CODED VARIABLES ACTUAL VALUE Y S/N Time Mass of Time Mass of Conc. (Min) Adsorbent (Min) Adsorbent (mg/l) X1 X2 X1 X2 1 -K 0 5 1.5 1770.37 2 K 0 60 1.5 1297.32 3 0 -K 32.5 0.5 1462.89 4 0 K 32.5 2.5 1230.66 5 -1 -1 13.1 0.793 1696.55 6 -1 1 13.1 2.21 1591.90 7 1 -1 51.95 0.793 1417.02 8 1 1 51.95 2.21 1247.15 9 0 0 32.5 1.5 1340.32 10 0 0 32.5 1.5 1340.32 11 0 0 32.5 1.5 1340.32 12 0 0 32.5 1.5 1340.32 11

3.2

Analysis of Box-Wilson design

Batch experiments were carried out randomly at five different levels: extreme (k), intermediate (-1), null/centre point (0), intermediate (+1), and extreme (+K) with reference to two independent factors: stirring time (X1) and mass of adsorbent (X2). The dependent variable is the adsorbed concentration (Y). The value of K, from Equation (5) is: [ ]

= 1.414

(7)

Coded values of the process variables and their variation intervals are as shown in Table 10. Table 10: Coded values of Process variables and their variation intervals
Variable Symbol -1.414 Stirring time Mass of Adsorbent, g Concentration, mg/l X1 X2 Y 5.0 0.500 100 -1 13.0546 0.793 158.5786 0 32.500 1.500 300.00 Levels +1 51.9454 2.210
441.4214

+1.414 60.000 2.500 500.00

Variation interval 19.4454 0.707 141.4214

With the adsorption results, surface response studies were conducted. The total number of design points N for the Box-Wilson DoE was determined using Equation (6):

Thus, as depicted in Tables 2 4, we have a 22 factorial experimental design (k = 2), with four axial points (-1.414, 0, +1.414, 0) and four replicates of the central/null points ( ),

totaling 12 experiments. Table 5 shows the experimental plan (coded and actual values) for the 12 experiments, with replications (as illustrated in Tables 6 - 9) to establish the homogeneity of the experimental method. 3.2.1 Mathematical Modeling of Box-Wilson Design Using 3rd Order Polynomial The statistical model for Box-Wilson experimental design with a third order polynomial, having two factors (k = 2) could be expressed as:

(8)

Using the actual values, the regression coefficients are calculated as shown in Table 11.
12

Table 11: Estimated Regression Coefficients for the various Experiments


Regression Coefficient Experiment ZW1(5%H2SO4) ZW1(H2O) ZW2(5%H2SO4) ZW2(H2O) 812.8 743 741.8 776 36.9 45 97.7 47 1470.5 1489.6 1626.5 1521.6 -13.3 -40 -89 -44 -1.5 -1.3 -2.5 -1.3 -756.9 -656.9 -759.4 -684.9 -0.15 0.14 0.46 0.15 5.98 9.8 17.1 11.1 0.02 0.014 0.02 0.014 65.9 77.8 84.1 79.3

Substituting the estimated regression coefficients into Equation (8), we obtain the mathematical models for each of the experimental data as follows: I. Model 1: For the adsorbent washed with 5%H2SO4 (ZW1):

(9) II. Model 2: For the adsorbent washed with H2O (ZW1):

(10) III. Model 3: For the adsorbent washed with 5%H2SO4 (ZW2):

(11) IV. Model 4: For the adsorbent washed with H2O (ZW2):

(12) 3.3.2 Analysis of Variance (ANOVA) The goodness of fit is conducted by correlating the observed and predicted values. The correlation coefficient of multiple determination and its square (coefficient of multiple

determination) are calculated by the expressions (Montgomery, 2005):

13

where

= the error sum of squares,

= sum of square of predicted values (

).

Equations (9) - (12) are used to predict the surface responses ( ) to the factors X1 and X2.Then, the error sum of squares (or residual sum of square) is calculated using the expression:

The experimental error variance (i.e. the mean error sum of squares)

is obtained by:

where

= degree of freedom for error variance =

, (N = number of experimental

runs/trial = 12,

= number of coefficients in the model = 10). is then calculated by the following equation:

The estimated variance of coefficient

The F-value is calculated by the expression:

The significant of effects, which is also the significant coefficient, is estimated by comparing the values of the ratio confidence level( 4.0 with the critical value of the F-distribution at 95% , the effect is significant.

). If the ratio

RESULTS AND DISCUSSION

The predicted yield ( ) obtained from Equations (9) (12) for the different experiments their deviations from the experimental values ( ,the error sum of squares and variance are

shown in Tables 12 - 15.The results of the correlation coefficient of multiple determination and its square (coefficient of multiple determination) calculated from Equation (13)

and (14) are given in Table 16. The results showed that the model fitted with the
14

experimental values in the range of the variables investigated. Multiple regression analysis results indicated a high degree of correlation between the experimental values and those predicted by the model. The variance analysis of variable effects is shown on Tables 17 - 20.
Table 12: Predicted Yield from Model 1for Sorbent washed with 5%H2SO4 at T = 32oC (ZW1) CODED VARIABLES S/N Time Mass of (Min) Adsorbent X1 X2 1 -1.414 0 2 1.414 0 3 0 -1.414 4 0 1.414 5 -1 -1 6 -1 1 7 1 -1 8 1 1 9 0 0 10 0 0 11 0 0 12 0 0

ACTUAL VALUE Time Mass of (Min) Adsorbent X1 X2 5 1.5 60 1.5 32.5 0.5 32.5 2.5 13.1 0.793 13.1 2.21 51.95 0.793 51.95 2.21 32.5 1.5 32.5 1.5 32.5 1.5 32.5 1.5

Conc. (mg/l) 1687.95 1218.48 1435.29 828.74 1664.29 1257.83 1338.17 633.06 1364.51 1364.51 1364.51 1364.51

Conc. (mg/l) 1650.59 1652.24 1474.56 880.33 1706 1297.34 1669.97 980.57 1443.44 1443.44 1443.44 1443.44 1395.77 188147.7 1542.1 2661.53 1739.7 1561 110091.2 120763.2 6229.95 6229.95 6229.95 6229.95

Table 13: Predicted Yield from Model 2 for Sorbent washed with Water at T = 32oC (ZW1) CODED VARIABLES S/N Time Mass of (Min) Adsorbent X1 X2 1 -1.414 0 2 1.414 0 3 0 -1.414 4 0 1.414 5 -1 -1 6 -1 1 7 1 -1 8 1 1 9 0 0 10 0 0 11 0 0 12 0 0 ACTUAL VALUE Time Mass of (Min) Adsorbent X1 X2 5 1.5 60 1.5 32.5 0.5 32.5 2.5 13.1 0.793 13.1 2.21 51.95 0.793 51.95 2.21 32.5 1.5 32.5 1.5 32.5 1.5 32.5 1.5 Conc. (mg/l) 1759.62 1318.82 1416.30 1236.40 1650.03 1598.35 1382.75 1226.36 1332.22 1332.22 1332.22 1332.22 Conc. (mg/l) 1771.7 1284.95 1406.83 1257.28 1632.09 1586.37 1314.03 1187.73 1320.36 1320.36 1320.36 1320.36 145.93 1147.18 89.68 435.97 321.84 143.52 4722.43 1492.28 140.66 140.66 140.66 140.66

15

Table 14: Predicted Yield from Model 3 for Sorbent washed with 5%H2SO4at T = 32oC (ZW2) CODED VARIABLES S/N Time Mass of (Min) Adsorbent X1 X2 1 -1.414 0 2 1.414 0 3 0 -1.414 4 0 1.414 5 -1 -1 6 -1 1 7 1 -1 8 1 1 9 0 0 10 0 0 11 0 0 12 0 0 ACTUAL VALUE Time Mass of (Min) Adsorbent X1 X2 5 1.5 60 1.5 32.5 0.5 32.5 2.5 13.1 0.793 13.1 2.21 51.95 0.793 51.95 2.21 32.5 1.5 32.5 1.5 32.5 1.5 32.5 1.5

Conc. (mg/l) 1698.70 566.23 1644.22 151.24 1812.00 1164.72 1218.48 129.02 759.04 759.04 759.04 759.04

Conc. (mg/l) 1727.36 -278.76 1532.53 53.93 1770.72 1122.86 605.20 -467.97 618.43 618.43 618.43 618.43 821.40 714008.10 12474.66 9469.24 1704.04 1752.26 376112.36 356397.06 19771.17 19771.17 19771.17 19771.17

Table 15: Predicted Yield from Model 4 for Sorbent washed with Water at T = 32oC (ZW2) CODED VARIABLES ACTUAL VALUE Conc. Conc. S/N Time Mass of Time Mass of (mg/l) (mg/l) (Min) Adsorbent (Min) Adsorbent X1 X2 X1 X2 1 -1.414 0 5 1.5 1770.37 1788.86 341.88 2 1.414 0 60 1.5 1297.32 1296.61 0.50 3 0 -1.414 32.5 0.5 1462.89 1464.56 2.79 4 0 1.414 32.5 2.5 1230.66 1247.69 290.02 5 -1 -1 13.1 0.793 1696.55 1669.85 712.89 6 -1 1 13.1 2.21 1591.90 1565.57 693.27 7 1 -1 51.95 0.793 1417.02 1358.05 3477.46 8 1 1 51.95 2.21 1247.15 1203.72 1886.16 9 0 0 32.5 1.5 1340.32 1323.43 285.27 10 0 0 32.5 1.5 1340.32 1323.43 285.27 11 0 0 32.5 1.5 1340.32 1323.43 285.27 12 0 0 32.5 1.5 1340.32 1323.43 285.27

16

Table 16: Calculated values of Correlation Coefficients Experiment 45288.03 226411.08 25081573.08 Adsorbent washed with

0.99 0.999 0.97 0.999

0.99 0.999 0.94 0.999

5%H2SO4 (ZW1) Adsorbent washed with water (ZW1) Adsorbent washed with 5%H2SO4 (ZW2) Adsorbent washed with water (ZW2)

9061.47 1551823.79

4530.74 775911.90 4273.03

23641008.09 11924385.87 24121401.87

Table 17: Variance Analysis of Variable Effects for Model 1 Effect


15703.325 31.026 40352.563 34280579.227 118 86770312.154 150673.576 93050289024.849 546.012 812.8 36.9 1470.5 -13.3 -1.5 -756.9 -0.15 5.98 0.02 65.9 14.418 7297.463 5.611 0.0066 1918.738 0.0026 1.503 0.000002 414.663 94.438 296.318 31.527 340.669 298.580 8.623 23.798 164.392 10.473 S S S S S S S S S

F-Value ( )

Table 18: Variance Analysis of Variable Effects for Model 2


Effect 15703.325 31.026 40352.563 34280579.227 118 86770312.154 150673.576 93050289024.849 546.012 743 45 1489.6 -40 -1.3 -656.9 0.14 9.8 0.014 77.8 0.289 146.030 0.112 0.00013 38.396 0.00005 0.030 0.000000049 8.298 F-Value ( ) 7018.552 15194.835 14250.233 12786.913 11238.579 375.369 3193.891 4025.359 729.444 S S S S S S S S S

17

Table 19: Variance Analysis of Variable Effects for Model 3 Effect 741.8 15703.325 31.026 40352.563 34280579.227 118 86770312.154 150673.576 93050289024.849 546.012 97.7 1626.5 -89 -2.5 -759.4 0.46 17.1 0.02 84.1 49.411 25008.44 19.228 0.023 6575.525 0.0089 5.150 0.000008 1421.052 193.183 105.784 411.944 276.131 87.702 23.663 56.783 47.969 4.977 S S S S S S S S S F-Value ( )

Table 20: Variance Analysis of Variable Effects for Model 4 Effect


776

F-Value ( )

15703.325 31.026 40352.563 34280579.227 118 86770312.154 150673.576 93050289024.849 546.012

47 1521.6 -44 -1.3 -684.9


0.15 11.1 0.014 79.3

0.272 137.724 0.106 0.000125 36.212 0.000049 0.028 0.000000046 7.826

8118.044 16810.895 18282.708 13558.102 12953.896 456.896 4344.573 4268.132 803.549

S S S S S S S S S

From the analysis of variance for adsorbent washed with 5%H2SO4(ZW1), comparison of the F-value with the critical value of F-distribution at 95% confidence level as shown in Table 17 indicates that all the effects are significant.

18

Table 18 indicates that all the effects on the adsorbent washed with water (ZW1) are significant. The analysis of variance as shown in Table 19 for the adsorbent washed with 5%H2SO4 (ZW2) indicates that all the effects are significant. The analysis of variance as shown in Table 20 for the adsorbent washed with water (ZW2) indicates that all the effects are significant. The optimum operating conditions are determined using the following Matlab programs: Model 1: For the adsorbent washed with 5%H2SO4 (ZW1):
function y = myfun2(X) y = -(812.8 + 36.9*X(1)+1470.5*X(2)-(13.3*X(1)* X(2))(1.5*(X(1)^2))-(756.9*(X(2)^2))-(0.15*(X(1)^2)* X(2))+ (5.98*(X(1)*(X(2)^2))+ (0.02*(X(1)^3))+ (65.9*X(2)^3)));

X0 = [1.0, 1.0]; [X,fval] = fminunc(@myfun2,X0)

Model 2: For the adsorbent washed with H2O (ZW1):


function y = myfun4(X) y = -(743+(45*X(1))+(1489.6*X(2))-(40*X(1)*X(2))-(1.3*(X(1)^2))(656.9*(X(2)^2))+(0.14*(X(1)^2)*X(2))+(9.8*X(1)*(X(2)^2))+(0.014*(X( 1)^3))+(77.8*(X(2)^3)));

X0 = [1,1]; [X,fval] = fminunc(@myfun4,X0) Model 3: For the adsorbent washed with 5%H2SO4 (ZW2):
function y = myfun5(X) y = -((741.8)+(97.7*X(1))+(1626.5*X(2))-(89*X(1)*X(2))(2.5*(X(1)^2))(759.4*(X(2)^2))+(0.46*(X(1)^2)*X(2))+(17.1*X(1)*(X(2)^2))+(0.02*(X( 1)^3))+(84.1*(X(2)^3)));

X0 = [2,0.5]; % 0r X0 = [4,0.5]; [X,fval] = fminunc(@myfun5,X0)

Model 4: For the adsorbent washed with H2O (ZW2):


function y = myfun6(X) y = -(776+(47*X(1))+(1521.6*X(2))-(44*X(1)*X(2))-(1.3*(X(1)^2))(684.9*(X(2)^2))+(0.15*(X(1)^2)*X(2))+(11.1*X(1)*(X(2)^2))+(0.014*(X (1)^3))+(79.3*(X(2)^3)));

X0 = [5,0.5]; [X,fval] = fminunc(@myfun6,X0) 19

The optimum conditions for the four different experiments as obtained from the Matlab programs are shown in Table 21. Table 21: optimum conditions for the four Adsorption Experiments Experiment Time X1 (min) Adsorbent washed with 5%H2SO4 (ZW1) Adsorbent washed with water (ZW1) Adsorbent washed with 5%H2SO4 (ZW2) Adsorbent washed with water (ZW2) 11.02 3.37 7.25e17 3.23 Adsorbent Mass X2 (g) 1.13 1.52 -5.15e16 1.45 - 165.59e50 -1794.37 Yield Y (mg/l)

5.0

CONCLUSION

Experimental design is a vital tool used in process industrial, due to the diverse approaches available and the flexibility of the theory. It application allows a reliable estimation of the individual effects of factors, as well as the optimization of a process. In this study experimental design approach based on Box-Wilson design was used to investigate the effects of stirring time and mass of adsorbent on the yield response of adsorbents that were washed with two different liquids (5%H2SO4 and water) at temperature of 32oC. It was demonstrated, by applying the lack-of-fit concept, that the Box-Wilson experimental design was appropriate for estimating the coefficients in a full quadratic regression with two independent variables. Multiple regression analysis results indicated a high degree of correlation between the experimental values and those predicted by the model. The developed regression models adequately described the observed data. The statistical significance of the factors and their interactions were articulated and were found significant at 95% confidence level. The optimum conditions for the adsorption process were also studied. REFERENCES

20

1. Box, G. E. P., Hunter, J. S., Hunter W. G., (2005), Statistics for Experimenters. Design, Innovation, and Discovery. 2nd Ed. Hoboken, NJ: John Wiley & Sons, Inc. 2. Fisher, R. A., (1935), The Design of Experiments, New York, NY: Hafner Press. 3. Lazic, R. Z., (2004), Design of Experiments in Chemical Engineering, WILEYVCHVerlag GmbH & Co., KGaA, Germany. 4. Mason, R. L., Gunst, R. F., Hess, J. L., (2003), Statistical Design and Analysis of Experiments with Applications to Engineering and Science, 2nd Ed., John Wiley & Sons, USA. 5. Matthews, P. G., (2005), Design of Experiments with MINITAB. Milwaukee, WI: American Society for Quality. 6. Meyer, R. H., Montgomery, D. C., (2002), Response Surface Methodology: Process and Product Optimization using Designed Experiment, 2nd Ed., John Wiley & Sons, USA. 7. Montgomery, C. (2005), Design and Analysis of Experiments, 6th Ed., John Wiley & Sons, USA. 8. NIST/SEMATECH, e-Handbook of http:/www.itl.nist.gov/div898/handbook, 23/03/2011. Statistics Methods,

9. Offurum, J. C., Chukwu, M., M., (2011), Factorial Design for Adsorption of Solid Particles in Coal Effluent using Feather-Driven Adsorbent, Continental J. Applied Sciences, Vol. 6 (3),pp. 1-7. 10. Vahid, H., Mohmad, K., Alireza, Momadrezaee, (2011), An Experimental Design Approach to Determine Effect of the Operating Parameters on the Rate of RU promoted Ir Carbonylation of Methanol, World Academy of Science, Engineering and Technology, Vol. 73, pp.598-603.

21

You might also like