Continuous models differential approach

In the case of selective neutrality—this means that all variants have the same selective values—evolution can be modeled successfully by diffusion models. This approach is based on the analysis of partial differential equations that describe free diffusion in a continuous model of the sequence space. The results obtained thereby and their consequences for molecular evolution were recently reviewed by Kimura [2]. Differences in selective values were found to be prohibitive, at least until now, for an exact solution of the diffusion approach. Needless to say, no exact results are available for value landscapes as complicated as those discussed in Section IV.3. Approximations are available for special cases only. In particular, the assumption of rare mutations has to be made almost in every case, and this contradicts the strategy basic to the quasi-species model. 

This approach, the stage model, is simple because the equations are first-order ordinary differential ones that can be easily solved by Gear s algorithm. However, it is less accurate than the continuous model, especially when dealing with columns... 

Continuous models are easy to implement and, therefore, have been widely used to describe the drying of deformable porous media (e.g., see Chapters 3 and 4, Volume 1 of this series). In these models, gel properties are expressed in terms of average parameters (or parameter functions) such as the diffusion coefficient, thermal diffusivity and Young s modulus, and partial differential equations for mass, heat and momentum balances are solved by discretization ( top-down approach). 

Over the past 40 years, numerous modeling studies have focused on the cellular slime mold Dictyostelium discoideum. This is because D. discoideum provides an experimentally accessible and relatively simple system for studying key developmental processes like chemotaxis, cell sorting, and complex pattern formation. Early studies adopted continuous models [77-80]. More recent studies used elegant hybrid approaches that combined CA models and partial differential equations to model 2D and 3D problems involving aggregation and self-organization ofD. discoideum [81-84]. 

A better alternative for tissue growth modehng is the hybrid discrete-continuous approach. Hybrid discrete-continuous models employ a discrete algorithm to simulate the dynamics of a cell population, while processes such as diffusion and consumption of GFs or nutrients are described with a continuous, deterministic component usually based on partial differential equations. 

Export processes are often more complicated than the expression given in Equation 7, for many chemicals can escape across the air/water interface (volatilize) or, in rapidly depositing environments, be buried for indeterminate periods in deep sediment beds. Still, the majority of environmental models are simply variations on the mass-balance theme expressed by Equation 7. Some codes solve Equation 7 directly for relatively large control volumes, that is, they operate on "compartment" or "box" models of the environment. Models of aquatic systems can also be phrased in terms of continuous space, as opposed to the "compartment" approach of discrete spatial zones. In this case, the partial differential equations (which arise, for example, by taking the limit of Equation 7 as the control volume goes to zero) can be solved by finite difference or finite element numerical integration techniques. 

In order to illustrate this approach, we next consider the optimization of an ammonia synthesis reactor. Formulation of the reactor optimization problem includes the discretized modeling equations for a packed bed reactor, along with the set of knot placement constraints. The following case study illustrates how a differential-algebraic problem can be optimized efficiently using (27). In addition, suitable accuracy of the ODE model can be obtained at the optimum by directly enforcing error restrictions and adaptively adding elements. Finally, bounds on the continuous state profiles can be enforced directly in the optimization problem. 

In this chapter we will keep the description of transport simpler than Fick s law, which would eventually lead to partial differential equations and thus to rather complex models. Instead of allowing the concentration of a chemical to change continuously in space, we assume that the concentration distribution exhibits some coarse structure. As an extreme, but often sufficient, approximation we go back to the example of phenanthrene in a lake and ask whether it would be adequate to describe the mass balance of phenanthrene by using just the average concentration in the lake, a value calculated by dividing the total phenanthrene mass in the lake by the lake volume. If the measured concentration in the lake at any location or depth would not deviate too much from the mean (say, less than 20%), then it may be reasonable to replace the complex three-dimensional concentration distribution of phenanthrene (which would never be adequately known anyway) by just one value, the average lake concentration. In other words, in this approach we would describe the lake as a well-mixed reactor and could then use the fairly simple mathematical equations which we have introduced in Section 12.4 (see Fig 12.7). The model which results from such an approach is called a one-box model. 

This approach of subdividing space into an increasing number of discrete pieces provides the basis for many numerical computer models (e.g., the so-called finite difference models) an example will be discussed in Chapter 23. Although these models are extremely powerful and convenient for the analysis of field data, they often conceal the basic principles which are responsible for a given result. Therefore, in the next chapter we will discuss models which are not only continuous in time, but also continuous along one or several space axes. In this context continuous in space means that the concentrations are given not only as steadily varying functions in time [QY)], but also as functions in space [C,(r,x) or C,(t,x,y,z)]. Such models lead to partial differential equations. A prominent example is Fick s second law (Eq. 18-14). 

The same example was solved using MINOPT (Rojnuckarin and Floudas, 1994) by treating the PFR model as a differential model. The required input files are shown in the MINOPT manual. Kokossis and Floudas (1990) applied the presented approach for large-scale systems in which the reactor network superstructure consisted of four CSTRs and four PFR units interconnected in all possible ways. Each PFR unit was approximated by a cascade of equal volume CSTRs (up to 200-300 CSTRs in testing the approximation). Complex reactions taking place in continuous and semibatch reactors were studied. It is important to emphasize that despite the complexity of the postulated superstructure, relatively simple structure solutions were obtained with the proposed algorithmic strategy. 

Ciano et al. (2006) have used a finite element approach to model a tubular cell 0.3 m long. The equations are available in Ciano et al. (2006). Table 7.2 shows the partial differential equations and the mesh characteristics. This model is computationally demanding and the equations have been solved by adopting an iterative procedure. Initial guess values for temperature and current density are assumed (current density is calculated by means of a lumped model, as the function of the average temperature and the cell voltage). Momentum equation and continuity equation are... 

The macroscopic approach, in which it is not taken into account what happens inside the cell in detail, but only an overall view of the system is described. In fact, the system is considered as a black box from the fluid dynamic point of view and then, it is assumed that the cell behaves a mixed tank reactor (the values of the variables only depend on time and not on the position since only one value of every variable describes all positions). This assumption allows simplifying directly all the set of partial differential equations to an easier set of differential equations, one for each model species. For the case of a continuous-operation electrochemical cell, the mass balances take the form shown in (4.5), where [.S, ]... 

Fixed Coordinate Approaches. In the fixed coordinate approach to airshed modeling, the airshed is divided into a three-dimensional grid for the numerical solution of some form of (7), the specific form depending upon the simplifying assumptions made. We classify the general methods for solution of the continuity equations by conventional finite difference methods, particle in cell methods, and variational methods. Finite difference methods and particle in cell methods are discussed here. Variational methods involve assuming the form of the concentration distribution, usually in terms of an expansion of known functions, and evaluating coeflBcients in the expansion. There is currently active interest in the application of these techniques (23) however, they are not yet suflBciently well developed that they may be applied to the solution of three-dimensional time-dependent partial differential equations, such as (7). For this reason we will not discuss these methods here. 

There are basically two ways of modeling a flow field the fluid is either treated as a collection of molecules or is considered to be continuous and indeflnitely divisible - continuum modeling. The former approach can be of deterministic or probabilistic modeling, while in the latter approach the velocity, density, pressure, etc. are aU deflned at every point in space and time, and the conservation of mass, momentum and energy lead to a set of nonlinear partial differential equations (Navier-Stokes). Fluid modeling classiflcation is depicted schematically in Fig. 1. 

---