An interactive teaching system for bond graph modeling and simulation in bioengineering.
At the university level, the teaching of bioengineering involves a variety of topics ranging from biology, biochemistry, medicine, automation, and informatics. Therefore, training in this domain is a difficult problem (Akpan, 2001; Djordjevic, Gerla, Huptych, Lhotska, & Krajca, 2010; Gonzalez-Cruz, Rodriguez-Sotres, & Rodriguez-Penagos, 2003). Within the design process of a bioengineering course, significant problems are the modeling of bioprocesses and the simulation of these complex systems (Dochain, 2008; Petre, 2006; Roman, Sendrescu, Bobasu, Petre, & Popescu, 2011; Roman & Selisteanu, 2012; Selisteanu, Roman, & Sendrescu, 2010). Another important problem is the choice of teaching system to pass from theory to practice. The teaching material in the field of bioengineering can be classified as simulators, laboratory bioreactors, pilot bioreactors, and industrial bioreactors.
As was demonstrated by the large number of studies, the interactive computer simulations, designed to teach complex processes, have become very popular in all domains of science education, such as physics, chemistry, and biology (Holzinger, Kickmeier-Rust, Wassertheurer, & Hessinger, 2009). Nowadays, technological developments such as computer simulations can implement more effective the so-called inquiry learning (de Jong, 2006). By using simulations to model a phenomenon, students can perform experiments by changing some variables and then observe the effects of their changes; thus, they discover the properties of the original model (de Jong, 2006). Simulations as learning tools are engaging and can be valuable for bioengineering. On the whole, a major advantage of learning with interactive simulations can be seen in the highly constructivist nature of such learning processes (Gulbahar, Madran, & Kalelioglu, 2010; Holzinger et al., 2009).
Several works, such as (de Jong, 2006; Keselman, 2003; Lawson, 2002; Manlove, Lazonder, & de Jong, 2006), reported that the students have significant problems with different inquiry learning processes: they have difficulty choosing the right variables to work with; they do not necessarily draw the right conclusions from experiments; they may have difficulty linking experimental data and hypotheses; they fail to make predictions; they make mistakes when interpreting data; and more. Therefore, research currently focuses on finding cognitive tools that help to surmount these problems and produce efficient learning situations. More precisely, computer environments can integrate these cognitive tools with the simulation (de Jong, 2006).
However, essential practical details can be lost if engineering teaching is reduced only to lectures and digital simulations, bypassing physical/biochemical experiments because of their relatively high costs (Precup, Preitl, Radac, Petriu, Dragos, & Tar, 2011). For that reason, most of universities try to purchase or to develop real bioreactors (laboratory or pilot reactors); in this way the laboratories can be achieved by using real bioprocesses. Of course, it is essential to apply the theory on real bioprocesses, but in most cases this method is expensive, and in some situations there are biological risks; therefore, in many cases simulators are used (Gonzalez-Cruz et al., 2003). Linked to this approach, in the last period a lot of interactive, computer and web-based teaching systems were reported, but especially related to physics/chemistry/biology education (Bunce, VandenPlas, & Havanki, 2006; Chang, Chen, Lin, & Sung, 2008; Gibbons, Evans, Payne, Shah, & Griffin, 2004; Limniou, Papadopoulos, & Whitehead, 2009, Milrad, 2002). Consequently, the use of software simulators is common and it constitutes an alternative to real bioreactors.
Concerning the choice of a particular teaching tool for biotechnology education at undergraduate and/or master level, it is obvious that this education requires intensive training in laboratory procedures. One of the key elements is the need to allow the student to familiarize laboratory techniques in balance with regular theory (Diwakar, Achuthan, Nedungadi, & Nair, 2011). The only way to achieve this goal with low costs is to use interactive, computer, and web- based teaching systems, including virtual labs. This tendency was reported in several works of the last decade, such as (Akpan; 2001; Gibbons et al., 2004; Gonzalez-Cruz et al., 2003; Palladino, 2002; Palladino & Cosentino, 2001; Smith & Emmeluth, 2002), and in numerous recent papers (Cid & Rajal; 2011; Diwakar et al., 2011; Roman et al., 2011).
The reasons to focus on interactive computer-based teaching systems for bioengineering are many (Diwakar et al., 2011). First, as already we outlined, the experimental setup cost is a big problem. Another motivation is the need to introduce computer-based teaching systems and virtual labs, which use mathematical modeling and computational methods. From the learning perspective, such systems allow students to manipulate advanced but common-to-use simulation tools. For biotechnology, a strong motivation for the shift to computer-based-and-virtual-lab paradigm, is the explosion of data-rich information sets, which are difficult to understand without the use of analytical tools. Another reason is the increasing interest in in silico experimentation due to ethical considerations, risk, and complications involved in human and animal research (Diwakar et al., 2011).
Combining theory will software simulators achieves good training results. In this paper, which is an extended work of (Roman et al., 2011), we present an interactive teaching system developed by using the 20-sim modeling and simulation software environment (registered trademark of Controllab Products B.V., Netherlands). This system uses friendly graphical user interfaces and comprises sets of different experiments. First, a set of experiments with 20-sim software package is designed; the final task is to achieve bond graph prototype simulators, which can be combined to obtain simulators for complex bioprocesses. Second, a set of modeling exercises is designed to achieve the mathematical models of these prototype bioprocesses. Every experiment consists in a short tutorial, the body of experiment and several questions and tasks for master students.
These sets of modeling and simulation experiments are grouped into a teaching system, which is implemented at the University of Craiova's Department of Automation, Electronics, and Mechatronics, within the Automation of Complex Systems master program (the Bioengineering course). At the University of Craiova, the curricula and structure of this master program are organized in the electrical-engineering domain, and it follows the Bologna process (see, for example, Martins, Thiriet, Bonnaud, Hoffmann, Robert, Benlloch et al., 2008).
The proposed teaching system can be seen as a pre-laboratory setup, in which we performed both an analysis of learning performance of this teaching system, which showed promising results, and assessed the learning technique.
The paper is organized as follows. First, we provide a short introduction in the field of bioprocesses modeling via the bond graph approach. Then, we present an overview of the teaching system. As well, we examine the implementation of some prototype bioprocess simulators in a 20-sim environment and present a simulator for two interconnected bioprocesses. Subsequently, this section deals with a set of modeling exercises, which help the students to construct the mathematical model of prototype bioprocesses, obtained from bond graph simulators. The next section deals with an analysis of the learning performance of proposed teaching system. The analysis is performed by using a group of master students from the bioengineering course. The subsequent section presents several results obtained by using the analysis of variance. Finally, we present some remarks.
Technical background: Bioprocess modeling issues and bond graph approach
In industry, the bioprocesses take place inside biological reactors (bioreactors) in which several biological reactions occur simultaneously in a liquid medium (Dochain, 2008; Petre, 2006). The bioreactors can operate in three modes: continuous, fed-batch, and batch mode. A fed-batch bioreactor (FBB) initially contains a small amount of substrates and micro-organisms and is progressively filled with the influent substrates. When the FBB is full, the content is harvested. A batch bioreactor is filled with the reactant mixture: substrates and micro-organisms and allows for a particular time period for the reactions inside the reactor; after some time the products are removed. In a continuous stirred tank bioreactor (CSTB), the substrates are fed to the bioreactor continuously and an effluent stream is continuously withdrawn such that the culture volume is constant.
The bioprocesses are characterized by several difficult issues such as strongly nonlinearity of kinetics, unavailability of cheap, online instrumentation, etc. Therefore, there are problems concerning the development of a unified modeling approach. However, even if the bioprocess modeling is a difficult task, by using the mass balance of components and obeying some modeling rules, we can obtain a dynamical state-space model (described using nonlinear differential equations) by using either classical modeling methods (Dochain, 2008) or the bond graph methodology (Roman & Selisteanu, 2012; Selisteanu et al., 2010). The bond graph method is a relative new approach for bioprocess modeling, but the results are quite encouraging, from both modeling and simulation points of view. Therefore, it is appropriate to implement teaching systems (especially for master and PhD studies) based on this approach.
The bond graph method uses the effort-flow analogy to describe physical processes. A bond graph consists of subsystems linked together by lines representing power bonds. Each process is described by a pair of variables, effort e and flow f, and their product is the power. The direction of power is depicted by a half arrow. In a dynamic system the effort and the flow varies and, therefore, the power fluctuates in time (Karnopp & Rosenberg, 1974; Thoma, 1975). A specific approach adapted to physical system particularities, the pseudo bond graph, is more suitable for the modeling of chemical and biochemical reactions due to the meaning of effort and flow variables involved whose products do not have the physical dimension of power (Dauphin-Tanguy, 2000; Thoma & Ould Bouamama, 2000). Pseudo bond graphs keep both the unitary characteristic and basic methodology benefits. Two other types of variables are important in describing dynamic systems, and these variables (energy variables) are the generalized momentum p and generalized displacement q (Karnopp & Rosenberg, 1974).
An advantage of bond graph method over other techniques is that models of various systems belonging to different domains can be expressed using a set of only nine elements: inertial elements (I), capacitive elements (C), resistive elements (R), effort sources (Se) and flow sources (Sf), transformer elements (TF) and gyrator elements (GY), effort junctions (J0), and flow junctions (J1). I, C, and R elements are passive elements because they convert the supplied energy into stored or dissipated energy. Se and Sf elements are active elements because they supply power to the system, and TF, GY, 0 and 1-junctions are junction elements that serve to connect I, C, R, Se, and Sf, and constitute the junction structure of bond graph model. The concept of causality is an important bond graph concept. Causality is represented on a model by causal stroke placed perpendicular to the bond at one of its ends, indicating the direction of effort variable. Causal stroke assignment is independent of the power flow direction. This leads to the description of bond graphs in the form of state-space equation. The sources (Se and Sf) have fixed causality, the dissipative element (R) has free causality depending on the causality of the other elements of bond graph, and the storage elements (I, C) have preferential causality (integral causality or derivative causality) but it is always desirable that C and I elements be in integral causality. Transformers, gyrators, and junction elements have constrainedly causality.
One of the simplest biological reactions is the micro-organism growth process (Dochain, 2008; Petre, 2006) with the next reaction scheme:
S [??] X (1)
with S the substrate, X the biomass, and [phi] the reaction rate.
This growth reaction represents in fact a prototype reaction, which can be found in almost every bioprocess. The dynamic of the concentrations of components from scheme (1) can be modeled considering the mass balance of components. In the next section, we will achieve in 20-sim the bond graph models for two prototype bioprocesses: the first one for a bioprocess taking place into a batch bioreactor and the second one for a CSTB case. We will also develop a bond graph model of a bioprocess carried out in two interconnected bioreactors. These prototype simulators are built by the students in the frame of several experiments described in the following section.
Description of the teaching system
The proposed teaching system implements an interactive manner in experiments' progress. These experiments allow master students to learn about biochemical reactions, reaction schemes, kinetics, types of bioreactors, bioprocess models, bond graph methodology, etc. The exercises allow the students to modify the kinetic parameters, choose the desired type of bioreactor, plot time evolution of some biological variables (biomass, substrates, products), and compare the obtained results.
In order to help the students, the experiments comprise a short tutorial (in electronic form), which must be read on the beginning of an experiment, and a small quiz, useful for checking the students' knowledge after the experiment.
First, a set of experiments with 20-sim software package is designed; the final task is to obtain bond graph prototype simulators, which can be combined in order to obtain simulators for complex bioprocesses:
* Experiment no. 1: Basics of 20-sim modeling and simulation environment
* Experiment no. 2: Bond graph elements in 20-sim--implementation and connectivity
* Experiment no. 3: Prototype batch bioprocess simulator in 20-sim
* Experiment no. 4: Prototype continuous bioprocess simulator in 20-sim
* Experiment no. 5: Two interconnected bioprocesses simulator
Second, a set of modeling exercises is designed via bond graph, which comprises procedures for the achievement of mathematical models:
* Exercise no. 1: Mass balances and accumulation of species in bond graph terms
* Exercise no. 2: Constitutive equations and modeling of reaction rates
* Exercise no. 3: Mathematical model of the prototype batch bioprocess acquired via bond graph approach
* Exercise no. 4: Mathematical model of the prototype continuous bioprocess obtained via bond graph
* Exercise no. 5: Mathematical model of two interconnected bioprocesses
Simulation experiments: Prototype simulators
Twenty-sim is an advanced modeling and simulation program (Controllab Products B.V., Netherlands) that runs under Microsoft Windows. Using 20-sim, the behavior of systems, such as electric, mechanical, hydraulic, and chemical systems or any combination of these systems can be simulated. Twenty-sim fully supports graphical modeling, allowing the design of dynamic systems in an intuitive and user-friendly way. The 20-sim software package can be used by the students for the simulation of bioprocesses modeled by bond graphs. These simulators are organized in libraries that exploit the modularity of the method, which is very useful for the students, who will be able finally to understand and to simulate complex bioprocesses. Next, the core issues of experiments 3, 4 and 5 will be shortly presented.
Prototype simulators for batch and continuous bioprocesses
After they read short tutorials, the students start each experiment that consists in the implementation in 20-sim of bond graph elements (Sf, C, TF, etc.) by using libraries, tools and interactive menus. The prototype bond graph simulators are obtaining starting from the scheme (1) and by taking into consideration the bioprocess type.
In the case of the batch bioreactor (experiment no. 3), there is no influent into or effluent stream from bioreactor, and the biomass X is periodically collected. For the development of bond graph model of this bioprocess, the reaction scheme (1), and the mass transfer through the bioreactor are taken into account. The prototype is obtained and placed into a special implementation window of 20-sim environment (see Fig. 1). The directions of half arrows correspond to the run of reaction, going out from the substrate S towards biomass X. The mass balances of the components involved in the bioreactor are represented by 0-junctions. Thus, we will have two 0-junctions corresponding to substrate S and biomass X. The accumulation of components S and X in the bioreactor is represented by bonds 1 and 7 and are modeled using capacitive elements C. For the modeling of yield coefficients, TF elements were used (Roman & Selisteanu, 2012). A difficult task is the modeling of reaction kinetics. The form of kinetics is complex, nonlinear and, in many cases, unknown. A general assumption is that a reaction can take place only if all reactants are presented in the bioreactor. Therefore, the reaction rates are necessarily zero whenever the concentration of one of the reactants is zero. In order to model the rate of reaction cp, a modulated two-port R element, denoted MR4> was used.
For the case of continuous bioprocess (experiment no. 4), the substrate is fed to CSTB continuously, and an effluent stream is continuously withdrawn such that the culture volume is constant. From the reaction scheme (1) and taking into account the mass transfer, using the bond graph modeling procedure, the pseudo bond graph model is achieved and is given in Figure 2. The mass balances of components are represented by two 0-junctions: [0.sub.1,2,3,4] (mass balance for S), and [0.sub.8,9,10] (mass balance for biomass X).
A modulated two-port resistive element MR6,7 was used to model the kinetics. Mass flow of the component entering the reaction is modeled using a source flow element [Sf.sub.1]. The output flows of the reaction components are modeled by using flow sources represented by bonds 3 and 10: [S.sub.f3], [Sf.sub.10]. The accumulation of substrate and biomass in CSTB is represented by bonds 2 and 9, and they are modeled using capacitive elements C (Roman & Selisteanu, 2012).
After the implementation of bond graphs, the students can use all the facilities of 20-sim environment to add or to remove elements, to run simulations, to modify various parameters, to obtain and to analyze the evolution of biological variables (also, the dynamical state-space models can be obtained according to the procedures described in next subsections).
A simulator for two interconnected bioprocesses
In order to exploit the modularity of bond graph approach, embedded in the 20-sim package, the students can realize more complex simulators by using the prototypes. Next, an example of such simulator is shortly described (experiment no. 5).
The activated sludge bioprocess is an aerobic process of biological wastewater treatment (Dochain, 2008; Roman & Selisteanu, 2012). In practice, this bioprocess takes place inside CSTBs or in the so-called sequencing batch reactors. Usually, the activated sludge bioprocesses operate in at least two interconnected tanks, as in Figure 3: an aerator in which degradation of pollutants takes place and a sedimentation tank (settler) in which the liquid is clarified. That is, the biomass is separated from the treated wastewater. Part of the settled biomass is fed back to bioreactor, while the surplus biomass is removed. The reaction may be described by a simple autocatalytic aerobic microbial growth that can be represented by the reaction scheme:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
where S, X and C are respectively the pollutant, biomass and dissolved oxygen, [phi] is the reaction rate and [k.sub.1] and [k.sub.2] are yield coefficients.
In bond graph terms, the mass balances of components involved in the aerator are represented by three 0-junctions: [0.sub.1,2,3,4] (mass balance for S), [0.sub.6,7,8,9] (mass balance for C), and [0.sub.12,13,14] (mass balance for X), and the mass balance of component involved in the settler is given by one 0-junction: [0.sub.16,17,18]. The accumulations of species S, C, and X in bioreactor are represented by bonds 3, 8, 14, and 17, they are modeled using capacitive elements C. For the modeling of the reaction rate a two-port modulated R element, [MR.sub.11,12], was used. Mass flows of the components entering the reaction are modeled using flow source elements [Sf.sub.1] and [Sf.sub.6]. Also, the mass flow of recycled biomass is modeled using [Sf.sub.16]. The transformer elements [TF.sub.4,5], and [TF.sub.9,10] were introduced to model the yield coefficients. The output flows of the components exiting from the reaction are modeled using flow sources elements Sf represented by bonds 2, 7, 15, 18. The output flow of component X from Aerator is an input flow for Settler (Roman & Selisteanu, 2012).
The simulator presented in Figure 4 can be implemented by the students either by using the whole bond graph stepby-step procedure or by connecting the corresponding libraries for two continuous prototype processes (as in Fig. 2) interconnected into a cascade structure as in the schema from Figure 3.
Bond graph models of prototype bioprocesses and modeling exercises
In the second set of exercises, students can use the simulators obtained in the first set to develop dynamical models of bioprocesses. The exercises structure is the same as in the case of 20-sim experiments: short tutorials, main, and questions. The mathematical models are obtained via bond graph by writing the characteristic equations for both elements and junction structure and taking into account the constructive and process characteristics in mathematical terms. All the ingredients required for the design of mathematical models are studied by the students within exercises 1 and 2. After that, in the exercises 3, 4, and 5 the students learn how to build the mathematical models for the prototype bioprocesses and the activated sludge bioprocess (interconnected bioprocesses), respectively. To illustrate the procedure, in the following the main body of exercises 3 and 4 will be shortly described.
The mathematical model of batch bioprocess
By using the next procedure, the students obtain the model of batch bioprocess (in the frame of exercise 3).
(i) The constitutive equations of C-elements are as follows:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (3)
(ii) From the constitutive relations of 1-junction ([1.sub.3,4]) and MR element, we obtain: [f.sub.3] = [f.sub.4], [f.sub.4] being proportional to the reaction rate [phi] and V.
(iii) Using the constitutive relations of transformer elements and taking into account the signification of bond graph elements, the dynamical model is obtained:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (4)
(iv) The model (4) expresses the equations of mass balance for (1). The dynamical behavior of the concentrations can be easily obtained from the system (4):
dS/dt = -[phi], dX/dt = [phi]. (5)
Even if this system seems to be a simple one, the kinetics given by reaction rate [phi] can be very complex. The behavior of mathematical model (in the form (4) or in the form of concentrations' dynamics (5)) is studied by the students, who can use the 20-sim facilities to plot and to analyze the time evolution of main biological variables. As an example, in Figure 5, the profiles of biomass and substrate concentrations are presented as they are obtained from 20-sim. The students can observe the typical behavior of this process: the consumption of substrate associated with biomass production. They can change the form of kinetics, the values of some parameters, and the solver of differential equations (the type of solver can be crucial for stiff systems. See, for example, Figure 6).
The mathematical model of continuous bioprocess
In the frame of exercise 4, the students obtain the mathematical model of continuous bioprocess, by using the same modeling procedure as in the batch case.
(i) The accumulations of substrate and biomass represented by bonds 2 and 9, and modeled using capacitive elements C, give the following constitutive relations:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (6)
(ii) The significance of bond graph elements is as follows: [e.sub.2] is the substrate concentration S (g/l), [e.sub.9] the biomass concentration X (g/l), [f.sub.6] is [phi] x V, [C.sub.2] = [C.sub.9] = V (1) is the reactor volume, [Sf.sub.3] = [Sf.sub.10] = [F.sub.0], with [F.sub.0] the output flow (l/h), and [f.sub.1] = [F.sub.in] [S.sub.in], where [F.sub.in] is the influent substrate flow (l/h) and [S.sub.in] the influent substrate concentration (g/l).
(iii) Therefore, from (6) and taking into account the constitutive relations of junction elements, the dynamical model can be obtained:
V x dS/dt = [F.sub.in][S.sub.in] - [F.sub.0]S - [phi]V, V x dX/dt = -[F.sub.0]X + [phi]V. (7)
(iv) The model (7) expresses the equations of mass balance for (1). The dynamical behavior of concentrations can be easily obtained from (7). From the equation of continuity [F.sub.in] = [F.sub.0] and using the so-called dilution rate D = [F.sub.in] / V = 1 / [t.sub.r], where [t.sub.r] is the medium residence time, equations (7) become:
dS/dt = [DS.sub.in] - DS - [phi], Dx/dt = -DX + [phi] (8)
The behavior of mathematical model (in the form  or in the form of concentrations' dynamics ) is studied by the students. By using the 20-sim capabilities, the evolution of main biological variables can be analyzed. As an example, in Figure 7 the profiles of biomass and substrate concentrations are presented as they are obtained from 20-sim. The students can change the form of kinetics, the values of some biological and simulation parameters, etc.
Analysis of learning performance Participants and experimental condition
To evaluate the impact of the teaching system, we compared the learning performance by using two learning conditions: (i) a traditional theoretical set of lessons, which includes classical text descriptions, static images, mathematical issues, etc., and (ii) the proposed computer-based interactive teaching system that covers the same learning content. The so-called null hypothesis would be that the different learning conditions do not imply different learning performance (Holzinger et al., 2009). The comparison was achieved with a procedure based on pre-tests and post-tests. The learning performance was measured using a multiple-choice test. To accomplish this study, 44 master students from Automation of Complex Systems master program (second year of study, 2010-2011, bioengineering) at the University of Craiova were used. Thirty-one men and thirteen women participated.
To assess knowledge and learning performance, a pre-test/post-test scenario was used, with a multiple-choice knowledge test, including twenty questions about the modeling of bioprocesses. Each of these questions represented a significant aspect of the discipline and demanded a systematic understanding to ensure choosing the correct answer. Every question included five possible answers; checking none, one, or more alternatives could be correct. The minimum score was 0 and the maximum score was 20. The pre-test and post-test used the same questions, but presented in different order. Because the complexity of the full teaching system, the tests focused only on a pair of experiment/modeling exercise. For example, one of these pairs contains Experiment 3: "Prototype batch bioprocess simulator in 20-sim" and Exercise 3: "Mathematical model of the prototype batch bioprocess acquired via bond graph." These tests can be seen as a kind of pre-laboratory training, useful for the students in order to deal more easily with the real equipment, such as laboratory or pilot bioreactors (in our case, the bioengineering laboratory is equipped with a BIOSTAT Sartorius bioreactor [Nisipeanu, Bunciu, & Stanica, 2011]).
The students were requested to provide biographical data and to complete the pre-tests to assess the prior knowledge in bioprocess modeling field. Then, the students were randomly divided into two groups, one denoted T (for the first learning condition, traditional) and one by C (the second learning condition, computer-based). Each group consists of 22 students (group T: 17 men and 5 women, and group C: 14 men and 8 women). Group T was provided with hard copies of the lessons set, and the participants from group C used the PCs from the classroom, with full access to the interactive teaching system. Both groups were given the same amount of time to learn. After a short pause, the students were asked to complete the post-tests. The entire procedure took about 110 min, which is a standard time period.
The use of computer-based simulators as pre-laboratory training is in accordance with cognitive load theory and constructivism (Kirschner, 2001; Kirschner, 2002; Limniou et al, 2009; Shiland, 1999). The entire procedure implies that the master students: (a) do not receive a massive amount of information at the same time, (b) have enough time to handle the useful information, (c) construct their new information based on the previous knowledge, (d) confirm or reject their initial hypothesis, and (e) think on bond graph method related to biochemical phenomena.
Additionally, we distributed some questionnaires to the entire group of 44 students, providing a number of statements about traditional/computer-based teaching approach, their interest in bioengineering, understanding of theoretical and practical problems previously studied (pre-requirements), their interest in real/simulated experiments before and after bioengineering course and so on.
Results and discussion
First, the results concerning the pre-tests are presented in Table 1. The mean scores and standard deviations were calculated by using simple routines provided in Matlab (The MathWorks, Inc., USA). As was expected, no significant differences were found between the two groups (T and C). In group T, the mean score in pre-test was M = 11.63, and the standard deviation SD = 1.43. In group C, the results were: M = 11.95, SD = 1.64. As reported in various references, the gender issues related to learning conditions and learning performance are interesting (Caspi, Chajut, & Saporta, 2008; Franzoni & Assar, 2009). The results obtained in the frame of our pre-test show that the women from both T and C groups scored better than the men from the two groups. See Table 1.
However, the results obtained at the post-test were quite different with respect to pre-test (Table 2). For the group T, the mean score was M = 15.13, SD = 1.28, while for the group C we have M = 16.77, SD = 1.50. As expected, the results obtained by both groups are higher than in pre-tests. It is interesting that the results obtained by the men from group C are much better than the results obtained by the men from group T, which can mean that male students are more comfortable with computer-based training than they are with the traditional theoretical approach. As a conclusion, we can see that in the post-test, the group C resulted in a higher mean score than group T. Also, an interesting interaction of learning condition and gender was found for the post-tests.
In order to achieve a more appropriate statistical analysis, we apply the univariate analysis of variance (ANOVA), by using Statistics Toolbox, provided by Matlab. First, we computed the learning performance (i.e., the score in post-test compared with the score in pre-test), for both groups T and C. The results are presented in Figures 8 and 9, where the standard one-way ANOVA tables are given. The columns represent: SS (the sums of squares), df (degrees of freedom), MS (mean squares: SS/df), F statistic, and prob>F (the p value). ANOVA revealed a significant effect for both groups, (F = 72.85, p [approximately equal to] 0, and F = 102.32, p [approximately equal to] 0), as expected.
However, the repeated single factor analysis of variance is not quite suitable for this experiment because this kind of analysis can miss possible interaction effects. Because here we have a repeated measurements design, repeated measurements are taken of the same individuals, and it is necessary to establish a within-subject effect. More precisely, averaging over all individuals as if they were independent may give misleading results. Therefore, a two-way ANOVA was performed by using the Matlab environment. The results are plotted in Figure 10. The purpose of two-way ANOVA is to find out whether data from several groups have a common mean.
One-way ANOVA and two-way ANOVA differ in that the groups in two-way ANOVA have two categories of defining characteristics instead of one. In Figure 10, the columns in the source represent the effects related to the pre-test and post-test, and the rows represents the effects related to the learning condition (T and C groups). As can be observed, the p-value for the pre-test/post-test effect is zero (to four decimal places). This is a strong indication that the scores obtained by the students differ from the pre-test to the post-test. The p-value for the learning condition effect is 0.0026, which is also highly significant. This indicates that the C group is out-performing the T group in the obtained scores. There does not appear to be strong interaction between learning condition and test type.
In order to compare these results with another statistical computing environment, we also used the SPSS package (general linear model--repeated measurements). The tests of within-subjects contrasts are given in Figure 11. As you can see the results are similar. The confidence intervals computed with SPSS are presented in Figure 12.
The ANOVA can be performed if the homoscedasticity and normality of the data is checked. The data normality was tested by using an adaptation of the Kolmogorov-Smirnov test--more precisely, the Lilliefors test--provided by the Matlab. Also, the homogeneity of variance was checked by using the Breusch-Pagan test.
It is interesting to check whether the effect of learning method treatment depends on gender. This approach calls for a two-way factorial with repeated measures in a mixed model. However, this kind of analysis is more complex for our particular experiment, because the data concerning the gender are unbalanced for the considered groups. Therefore, the analysis for gender issues is based only on the descriptive statistics presented in Tables 1 and 2.
Concerning the assessment of learning technique, the results obtained from the questionnaires completed by the students are presented in Table 3. The highest interest is in real bioreactor experiments, but the difference between real experiments and simulators is quite small. Once the new methods of bioengineering training were used, the students' results were better, both during the semester and at the exam.
Nowadays, the advances in information technology provide new means for improvement of learning technologies. The presented computer-based teaching system allows an improvement in bioengineering education and training. The experiments touch on the important problems for bioprocesses. The system has didactic properties such as modularity of the package, friendly graphical user interfaces and so on.
This study confirmed the results of previous research on exploratory learning using interactive simulations. Still, it is necessary to offer additional assistance on the correct use of a simulation before beginning to learn with the computer-based teaching system. In future research, the evaluation of proposed teaching tool can be further improved. The implementation of the proposed bioengineering teaching system resulted in significantly higher learning performance than traditional learning conditions. The results pointed out that, with some exceptions, these effects are independent of individual learning strategies but not necessarily of gender.
As the use of real bioreactors is expensive and even dangerous, the exploitation of software simulators constitutes a good alternative. The interactive system helps master students to understand different bioprocess modeling and simulation issues. The package can be extended with new experiments, and can be seen as a pre-laboratory training system.
This work was supported by the grant POSDRU/89/1.5/S/61968, Project ID 61968 (2009), co-financed by the European Social Fund, 2007-2013.
Akpan, J. P. (2001). Issues associated with inserting computer simulations into biology instruction: A review of the literature. Electronic Journal of Science Education, 5(3).
Bunce, D. M., VandenPlas, J. R., & Havanki, K. L. (2006). Comparing the effectiveness on student achievement of a student response system versus online WebCT quizzes. Journal of Chemical Education, 83(3), 488-493.
Caspi, A., Chajut, E., & Saporta, K. (2008). Participation in class and in online discussions: Gender differences. Computers & Education, 50(3), 718-724.
Chang, K. E., Chen, Y. L., Lin, H. Y., & Sung, Y. T. (2008). Effects of learning support in simulation-based physics learning. Computers & Education, 51(4), 1486-1498.
Cid, A. G., & Rajal, V. B. (2011). New teaching strategies to improve student performance in fundamentals of biotechnology. Journal of Microbiology & Biology Education, 12(1), 1-2.
Dauphin-Tanguy, G. (2000). Les bond graphs. Paris: Hermes Sci. de Jong, T. (2006). Computer simulations: Technological advances in inquiry learning. Science, 312, 532-533.
Diwakar, S., Achuthan, K., Nedungadi, P., & Nair, B. (2011). Enhanced facilitation of biotechnology education in developing nations via virtual labs: Analysis, implementation and case-studies. International Journal of Computer Theory and Engineering, 3(1), 1793-8201.
Djordjevic, V., Gerla, V., Huptych, M., Lhotska, L., & Krajca, V. (2010). The development of modules for the support of education in the field of biomedical engineering. Electronics and Electrical Engineering, 6(102), 47-50.
Dochain, D. (Ed.). (2008). Automatic control of bioprocesses. London, UK: ISTE and John Wiley & Sons.
Franzoni, A.L., Assar, S. (2009). Student learning styles adaptation method based on teaching strategies and electronic media. Educational Technology & Society, 12(4), 15-29.
Gibbons, N. J., Evans, C., Payne, A., Shah, K., & Griffin, D. K. (2004). Computer simulations improve university instructional laboratories. Cell Biology Education, 3(4), 263-269.
Gonzalez-Cruz, J., Rodriguez-Sotres, R., & Rodriguez-Penagos, M. (2003). On the convenience of using a computer simulation to teach enzyme kinetics to undergraduate students with biological chemistry-related curricula. Biochemistry and Molecular Biology Education, 31(2), 93-101.
Giilbahar, Y., Madran, R. O., & Kalelioglu, F. (2010). Development and evaluation of an interactive WebQuest environment: "Web Macerasi." Educational Technology & Society, 13(3), 139-150.
Holzinger, A., Kickmeier-Rust, M. D., Wassertheurer, S., & Hessinger, M. (2009). Learning performance with interactive simulations in medical education: Lessons learned from results of learning complex physiological models with the HAEMOdynamics SIMulator. Computers & Education, 52(2), 292-301.
Karnopp, D., & Rosenberg, R. (1974). System dynamics: A unified approach. New York: John Wiley.
Keselman, A. (2003). Supporting inquiry learning by promoting normative understanding of multivariable causality. Journal of Research in Science Teaching, 40, 898-921.
Kirschner, P. A. (2001). Integrated electronic environments for collaboration teaching/learning. Research Dialogue in Learning and Instruction, 2, 1-9.
Kirschner, P. A. (2002). Cognitive load theory: Implications of cognitive load theory on the design of learning. Learning and Instruction, Guest Editorial, 12, 1-10.
Lawson, A. E. (2002). Sound and faulty arguments generated by pre-service biology teachers when testing hypotheses involving un-observable entities. Journal of Research in Science Teaching, 39, 237-252.
Limniou, M., Papadopoulos, N., & Whitehead, C. (2009). Integration of simulation into pre-laboratory chemical course: Computer cluster versus WebCT. Computers & Education 52(1), 45-52.
Manlove, S., Lazonder, A.W., & de Jong, T. (2006). Regulative support for collaborative scientific inquiry learning. Journal of Computer Assisted Learning, 22, 87-98.
Martins, M. J., Thiriet, J.-M., Bonnaud, O., Hoffmann, M., Robert, M., Benlloch, J., & Jervan, G. (2008). A survey of the evolution of the Bologna Process in EIE in Europe. Proceedings of 19th Annual Conference on Innovation in Education for Electrical and Information Engineering EAEEIE, Tallinn, Estonia (pp. 100-103). doi: 10.1109/EAEEIE.2008.4610166
Milrad, M. (2002). Using construction kits, modeling tools and system dynamics simulations to support collaborative discovery learning. Educational Technology & Society, 5(4), 76-87.
Nisipeanu, I. A., Bunciu E., & Stanica, R. (2011). Bioprocesses parameters control in the case of a BIOSTAT A PLUS bioreactor. Annals of the University of Craiova, Series: Automation, Computers, Electronics and Mechatronics, 8(2), 31-35.
Palladino, M. (2002). Learning about the human genome project via the web: Internet resources for biology students. American Biology Teacher, 64, 110-116.
Palladino, M., & Cosentino, E. (2001). A DNA fingerprinting simulation laboratory for biology students. American Biology Teacher, 63, 596-603.
Petre, E. (2006). A nonlinear adaptive controller for a fed-batch fermentation process. Control Engineering and Applied Informatics, 7(4), 31-40.
Precup, R. E., Preitl, S., Radac, M. B., Petriu, E. M., Dragos, C. A., & Tar, J. K. (2011). Experiment-based teaching in advanced control engineering. IEEE Transactions on Education, 54(3), 345-355.
Roman, M., & Selisteanu, D. (2012). Pseudo bond graph modelling of wastewater treatment bioprocesses. SIMULATION: Transactions of the Society for Modeling and Simulation International, 88(2), 233-251.
Roman, M., [section]endrescu, D., Bobasu, E., Petre, E., & Popescu, D. (2011). Teaching system for modelling and simulation of bioprocesses via bond graphs. Proceedings of 22nd Annual Conference on Innovation in Education for Electrical and Information Engineering EAEEIE, Maribor, Slovenia (pp. 192-199). Retrieved from http://eaeeie2011.uni-mb.si/eaeeie2011 _submission_17.pdf
Selisteanu, D., Roman, M., & Sendrescu, D. (2010). Pseudo bond graph modelling and on-line estimation of unknown kinetics for a wastewater biodegradation process. Simulation Modelling Practice and Theory, 18(9), 1297-1313.
Shiland, T. W. (1999). Constructivism: The implications for laboratory work. Journal of Chemical Education, 74, 730-731.
Smith, T., & Emmeluth, D. (2002). Introducing bioinformatics into the biology curriculum: Exploring the national center for biotechnology information. American Biology Teacher, 64, 93-99.
Thoma, J. (1975). Introduction to bond graphs and their applications. Oxford: Pergamon Press.
Thoma, J., & Ould Bouamama, B. (2000). Modelling and simulation in thermal and chemical engineering: A bond graph approach. Berlin, Germany: Springer-Verlag.
Monica Roman, Dorin Popescu and Dan Selisteanu *
Department of Automation, Electronics and Mechatronics, University of Craiova, 200585, Romania firstname.lastname@example.org // email@example.com // firstname.lastname@example.org
* Corresponding author
(Submitted September 17, 2012; Revised December 16, 2012; Accepted January 22, 2013)
Table 1. Results obtained in pre-test related to learning conditions and gender Test type Group Mean score M Standard deviation SD Pre-test T 11.6364 1.4325 C 11.9545 1.6469 T--women 12.8000 1.9235 T--men 11.2941 1.1048 C--women 12.1250 1.2464 C--men 11.8571 1.8752 Table 2. Results obtained in post-test related to learning conditions and gender Test type Group Mean score M Standard deviation SD Post-test T 15.1364 1.2834 C 16.7727 1.5097 T--women 16.2000 1.4832 T--men 14.8235 1.0744 C--women 16.7500 1.3887 C--men 16.7857 1.6257 Table 3. Assessment of learning technique and interest in bioengineering Before bioengineering After bioengineering course and experiments course and experiments High Medium Low High Medium Low Interest in 39% 54% 7% 74% 22% 4% bioengineering Understanding of 35% 44% 21% 73% 20% 7% theoretical problems previously studied Understanding of 29% 40% 31% 70% 24% 6% practical problems previously studied Interest in traditional 28% 53% 19% 32% 54% 14% theoretical methods Interest in 20-sim 40% 43% 17% 74% 21% 5% simulators Interest in real 71% 23% 6% 84% 14% 2% bioprocess experiments Figure 11. Two-way ANOVA results: Tests of within-subjects contrasts (SPSS) Tests of Within-Subjects Contrasts--SPSS Measure: MEASURE 1 Source Learning TestType Type III Sum df of Squares Learning Linear 21,011 1 Error (Learning) Linear 84,239 21 TestType Linear 380,557 1 Error (TestType) Linear 5,693 21 Learning * TestType Linear Linear 9,557 1 Error (Learning*TestType) Linear Linear 2,693 21 Source Mean Square F Sig. Learning 21,011 5,238 ,033 Error (Learning) 4,011 TestType 380,557 1403,731 ,000 Error (TestType) ,271 Learning * TestType 9,557 74,519 ,000 Error (Learning*TestType) ,128 Figure 12. Confidence intervals: results obtained with SPSS Test Value 95% Confidence Interval of the Difference t df Sig.(2- Mean tailed) Difference Lower Upper T--pre 38,102 21 ,000 11,63636 11,0012 12,2715 T--post 55,317 21 ,000 15,13636 14,5673 15,7054 C--pre 34,048 21 ,000 11,95455 11,2244 12,6847 C--post 52,110 21 ,000 16,77273 16,1034 17,4421
|Printer friendly Cite/link Email Feedback|
|Author:||Roman, Monica; Popescu, Dorin; Selisteanu, Dan|
|Publication:||Educational Technology & Society|
|Date:||Oct 1, 2013|
|Previous Article:||The effect of technology-supported co-sharing on L2 vocabulary strategy development.|
|Next Article:||Designing and implementing a personalized remedial learning system for enhancing the programming learning.|