Limitations due to Noise, Stability and Component Tolerance on the Solution of Partial Differential Equations by Differential Analysers

SIMULATION ◽  
1964 ◽  
Vol 3 (5) ◽  
pp. 45-52 ◽  
Author(s):  
Michael E. Fisher

The solution of partial differential equations by a differential analyser is considered with regard to the effects of noise, computational instability and the deviation of components from their ideal values. It is shown that the 'serial' method of solving parabolic, hyperbolic and elliptic equations leads to serious in stability which increases as the finite difference interval is reduced. The truncation error (due to the difference approximations) decreases as the interval is made smaller and consequently an 'optimal' ac curacy is reached when the unstable noise errors match the truncation errors. Evaluation shows that the attainable accuracy is severely limited, especially for hyperbolic and elliptic equations. The 'parallel' method is stable when applied to parabolic and hyperbolic (but not elliptic) equations and the attainable accuracy is then limited by the accumulation of component tolerances. Quantitative investigation shows how reasonably high accuracy can be achieved with a minimum of precise adjustments.

Author(s):  
Fabrizio Cuccu ◽  
Giovanni Porru

We investigate maxima and minima of some functionals associated with solutions to Dirichlet problems for elliptic equations. We prove existence results and, under suitable restrictions on the data, we show that any maximal configuration satisfies a special system of two equations. Next, we use the moving-plane method to find symmetry results for solutions of a system. We apply these results in our discussion of symmetry for the maximal configurations of the previous problem.


1971 ◽  
Vol 11 (03) ◽  
pp. 315-320 ◽  
Author(s):  
R.B. Lantz

Abstract Numerical diffusion (truncation error) can limit the usefulness of numerical finite-difference approximations to solve partial differential equations. Many reservoir simulation users are aware of these limitations but are not as familiar with actually quantifying the magnitude of the truncation error. This paper illustrates that, over a wide range of block size and time step, the truncation error expressions for convective-diffusion partial differential equations are quantitative. Since miscible, thermal, and immiscible processes can be of the convective-diffusion equation form, the truncation error expressions presented can provide guidelines for choosing block size-time step combinations that minimize the effect of numerical diffusion. Introduction Truncation error limits the use of numerical finite-difference approximations to solve partial differential equations. In the solution of convection-diffusion equations, such as occur in miscible displacement and thermal transport, truncation error results in an artificial dispersion term often denoted as numerical diffusion. The differential equations describing two-phase fluid flow can also be rearranged into a convection-diffusion form. And, in fact, miscible and immiscible differential equations have been shown to be completely analogous. In this form, it is easy to infer that numerical diffusion will result in an additional term resembling flow due to capillarity. Many users of numerical programs, and probably all numerical analysts, recognize that the magnitude of the numerical diffusivity for convection-diffusion equations can depend on both block size and time step. Most expressions developed in the literature have been used primarily to determine the order of the error rather than to quantify it. The primary purpose of this paper is to give the user more than just a qualitative feel for the importance of truncation error. In this paper, insofar as possible, analytical expressions for truncation error are compared by experiment to computed values for the numerical diffusivity. Consequently, the reservoir simulator user can observe that these expressions are quantitative and can use them as guidelines for choosing block sizes and time steps that keep the numerical diffusivity small. DEVELOPMENT OF EXPRESSIONS FOR TRUNCATION ERROR APPLICATION TO CONVECTION-DIFFUSION EQUATION To illustrate the method of quantifying numerical diffusivity, consider a convective-diffusion equation of the form: ..............(1) Symbols are defined in the Nomenclature. The first term on the right-hand side represents the diffusion, and the second term represents convection. Such an equation describes the flow of either a two-component miscible mixture or heat in one dimension with constant diffusivity. EXPLICIT DIFFERENCE FORMS An explicit expression for the truncation error (the space derivatives are approximated at a known time level) can be developed by examining the Taylor's series expansion representing first- and second-order derivatives. For the time derivative: .....(2) SPEJ P. 315


1970 ◽  
Vol 10 (02) ◽  
pp. 192-202 ◽  
Author(s):  
R.B. Lantz

Abstract In the past miscible displacement calculations have been approximated with two-phase reservoir simulators. Such calculations have neglected diffusional mixing between miscible components. In fact, no analog bas been proposed for rigorously treating miscible simulations with two-phase programs. This paper describes requirements that programs. This paper describes requirements that permit such a rigorous simulation. permit such a rigorous simulation. The sets of partial differential equations describing each of the displacement processes are shown to be exactly analogous if relative permeability and capillary pressure functions are permeability and capillary pressure functions are adjusted in a special manner. Application of the "miscible" analogy in a two-phase simulator, however, has several limitations, the most severe of which is the truncation error (numerical diffusion) typical of an immiscible formulation. Since this error is time-step and/or block-size dependent, numerical smearing can, in principle, be made as small as necessary. But this feature limits the practical applicability of the "miscible" analogy practical applicability of the "miscible" analogy to cases with rather large physical diffusion. The range of applicability and other limitations are outlined in the paper. Also, illustrative sample calculations are presented for linear, radial and layer-cake systems. Component densities and viscosities are varied in the linear model. Introduction In recent years, use of two- and three-phase reservoir simulators to calculate the performance of immiscible fluid displacement has become widespread. Reservoir simulators capable of calculating miscible displacement problems, however, have been limited to special use programs. The primary reason for this limitation has been the significant truncation error (numerical diffusion) typical of ordinary finite difference approximations to the miscible equations. The method of characteristics provided a means of making miscible displacement calculations without significant truncation error. A recently proposed second calculation technique, based on variational methods, also reduces numerical diffusion. Both of these calculational techniques can be used for immiscible calculations. Still, general miscible displacement applications such as gas cycling, enriched-gas injection, or tracer injection have historically required use of immiscible reservoir simulators for performance predictions. Larson et al. have reported an example of such use of a two-phase computer program. Displacement involving two components flowing within a single phase would appear to be analogous to a two-phase displacement. Yet, past miscible calculations using immiscible simulators made the two-phase saturation profile as near piston-like as possible and neglected component mixing due to possible and neglected component mixing due to diffusional processes. The capillary pressure function was chosen to minimize capillary flow. Also, in these miscible approximations, no provision had been made for viscosity variations provision had been made for viscosity variations with component concentration. Though mixing due to diffusional processes had been neglected, countercurrent diffusion due to component concentration differences in a miscible process should be essentially analogous to countercurrent capillary flow due to saturation differences in a two-phase system. This paper describes a method by which two- and three-phase reservoir simulators can be made to calculate miscible displacement rigorously. The only requirement of the method is that relative permeability and capillary pressure be special permeability and capillary pressure be special functions of saturation. With these properly chosen functions, the set of partial differential equations describing immiscible displacement becomes completely analogous to the partial differential equations describing miscible displacement. SPEJ P. 192


Sign in / Sign up

Export Citation Format

Share Document