You are on page 1of 4

2011 1st International Conference on Electric Power Equipment Switching Technology Xian China

Radiation and Nozzle-Ablation Models for CFD Simulations of Gas Circuit Breakers
Thomas Christen
ABB Switzerland Ltd, Corporate Research, Segelhofstrasse 1K, CH-5405 Baden-Dttwil, Switzerland

Abstract- CFD simulations of electric arcs in circuit breakers require physically reliable and computationally treatable models for the complex physics involved. This article discusses two entropy-production based concepts for radiation and nozzle ablation modeling with improved balance between physical foundation and needed computational resources.

I.

INTRODUCTION

be taken into account. Furthermore, turbulence models must be considered, which provide effective expressions for the viscosity and heat conductivity in mat, and je, respectively. Finally, initial and boundary conditions must be given. For more details see Ref. [3] in this volume. In the remainder, PE and the CFD boundary conditions at an irradiated insulating polymer surface will be discussed. III. RADIATION MODELING

The interruption performance of HV gas circuit breakers (GCB) depends on a number of complex physical phenomena associated with electric arc physics [1,2], which are still poorly understood in detail. Nevertheless, the computational fluid dynamics (CFD) based optimization of GCB design [3] for 3d geometries, which needs huge numerical resources, requires physically meaningful and numerically treatable models with reliable quantitative and qualitative predictive efficiency. This article focuses on radiation and ablation relevant for the high-current arcing phase. Those are complicated phenomena far from local thermal equilibrium (LTE). Section II briefly recalls the basic CFD context, and Sect. III and IV discuss the radiation model and the nozzle ablation model, respectively. II. CFD EQUATIONS

A. Radiative Transfer Heat radiation in the gas/plasma is described by the specific radiation intensity I(,x,t) (unit [W/m2 Hz sr]), which depends on frequency , direction unit vector , location x, and time t. It obeys the radiative transfer equation (RTE) c-1t I + I = (B - I ) , (4)

In the framework of an effective single-fluid gas or plasma [2] with known thermodynamic and caloric equations of states relating temperature T, pressure p, mass density , and specific energy e (or enthalpy h), the mass, momentum, and energy balance equations are , t + ( u) = , t ( u) + mat = f t ( etot) + je = PJ - PE ,
2

where c is the light speed, the absorption coefficient, B=2h 3n(eq)/c2 the radiation intensity of LTE matter, n(eq ) = (exp(h/kT)-1)-1 the Bose-Einstein distribution for LTE photons, and h and k are the Planck and Boltzmann constants. The RTE is a 6-d problem, and depends on p, T, and in a very complicated way on [6], which requires an appropriate modeling approach. While the net emission model [6] and the often used P1 model are oversimplifying, the discrete ordinate model (DOM) is computationally too expensive for exact realistic spatially 3d simulations. More suited is the so-called M1 model, which nevertheless has some severe drawbacks [5]. B. Moment expansion with entropy production closure In order to model radiation, we define moments of I E(x,t) F(x,t) (x,t) = = = , c-1 d d I c d d I , c-1 d d : I ,
-1

(1) (2) (3)

where u is the flow velocity, etot = e + u /2, mat is the stress tensor, je is the energy flux density, and t and denote time derivative and nabla symbol, respectively. The production terms on the right hand sides refer to effects like electrical Joule heating (PJ), heat radiation (PE), contact erosion (e.g. evaporation of ejected metal droplets, ; but usually =0), magnetic forces (f) etc. Additionally, the Maxwell equations for electric and magnetic fields in the (quasi-neutral) plasma described by an LTE electric conductivity and permittivity must

(5) (6) (7)

associated with energy density, flux density, and stress tensor, respectively (all units [J/m3] for convenience). Multiplying (4) by 1 and , respectively, and subsequent integrations with respect to and give c-1t E + F = E(E(eq)-E) c-1t F + = -F F , (8) (9)

978-1-4577-1272-2/$26.00 2011 IEEE

471

for the variables E and F, where E = 4SBT /c with the Stefan-Boltzmann constant SB. Mean absorption coefficients E and F are defined by obvious integrals containing I. The radiation stress-tensor can be expressed as [4,5]
(eq) 4

(1+2vj)/3 with j>1 (e.g., j=2) may reasonably approximate the VEF in practice.

kl= ( (1-)kl + (3-1)FkFl/F2)E ,

(10)

where kl is the Kronecker delta symbol and is the so-called Eddington factor. The quantities E, F, and are functionals of the unknown I(). A closure of the moment equations is a procedure that determines these quantities. Various mathematical issues and boundary conditions to E and F at solid surfaces can be found in [4,5]; here we mainly provide the receipt to determine E, F, and for given . Early work by Kohler (see [7]) motivates a closure by minimizing the entropy production rate [4,5]

s [I ] = k d d h-1 -1 ln(R) (B-I ) ,

(11)

Fig. 1. Effective absorption coefficients E (solid) and F (dashed-dotted) as functions of E/E(eq) for shown in the inset, at v 0, T = 10000 K. Planck mean 2.15/m (dotted); Rosseland mean 1.15/m (dashed). Large E limit: Min()=1.

where R=n(eq)(1+n)/n(1+n(eq)) and n=c2I /2h3, subject to the constraints defined by (5) and (6) with E and F kept constant at minimization. This constrained minimization provides an I that depends additionally on E and F. Hence E, F, and (variable Eddington factor,VEF) become functions of E and F. This closes (8) and (9) at an expense of being nonlinear. The concept is neither restricted to emission and absorption nor to two moments; scattering can be included and an arbitrary number of moments can be considered [5]. For some well-known limit cases this concept provides correct results, whereas other closures like P1 and M1 do not [5]. To show this, note first that the isotropy of implies that E, F, and can be expressed as functions of E and v= lF l/E, with 0 v 1, where v is an average radiation velocity in units of c. Consider first isotropic radiation (v=0) where =1/3. The case E=E(eq) refers to equilibrium radiation, where E and F are given by the so-called Rosseland mean [4,5] describing fully diffusive photons in an optically dense medium. In the emission limit, defined by E<<E(eq), E equals the so-called Planck mean [4,5]. Consider now the free streaming limit (fully directed radiation, v=1), where =1 (flux limiting) must hold. All these cases are reproduced by our approach; the results can be understood in terms of photon population [4,5]. C. Illustrative Results As an exemplary case, E and F are shown in Fig. 1 as functions of E/E(eq) at v=0 for a simplistic spectrum consisting of tree different absorption bands (inset). The result shows the cross-over from the emission limit for small E, where E equals the Planck mean, to the large E limit, where E and F converge to Min()=1, and both pass at E/E(eq) =1 (equilibrium) the Rosseland mean. For VEF results see [4,5]; we only mention that (v)=

Once the functions E(E,v), F(E,v), and (E,v), are known, the coupled set of equations (1)-(3) and (8)-(9) can be used for a complete GCB simulation. Due to power balance, PE = E(E(eq)-E) holds in (3). In typical cases, quasi-steady-state solutions of (8) and (9) are sufficient because of the fast propagation of radiation. IV. ABLATION MODELING

To solve the CFD equations, boundary conditions are needed. Radiation induced nozzle material ablation is important for the pressure build-up in the GCB, and it affects the CFD boundary conditions for (1)-(3) at the nozzle surface, which depend then on the impinging heat flux. For general breaker gas and solid materials, the partial pressure of the ablated material must be considered, and the gas temperature must be calculated by mixing of the injected with the already present mass [3]. This paper focuses on the determination of the hydrodynamic properties of the ablated vapor, and will not discuss mixing issues. A. Ablation boundary conditions The problem is sketched in Fig. 2 [8]. Part of the arc radiation removes material from the solid surface by combined thermal decomposition and photo-ablation. The degrees of freedom (translational, vibrational, rotational, electronic excitations, molar fraction of species, ...) of the ablated material equilibrate to LTE after passing a nonequilibrium (Knudsen) layer, in which another part of the radiation may still be absorbed. Since the vapor temperature will be below the ionization temperature of the compounds, it is not necessary to consider a separate plasma layer with different electron and ion temperatures. Note that solid polymers are meta-stable states; thus the
472

ablation process cannot be treated uniquely with material functions for full thermodynamic equilibrium. The properties depend rather on the real decomposition kinetics.

Fig. 2. Ablation boundary with non-LTE Knudsen layer between solid (at decomposition temperature Td) and LTE vapor (temperature Tb). During traversal through the non-LTE region, the various degrees of freedom with different temperatures (dashed) equilibrate towards a unique temperature T at the LTE boundary.

Fig. 3. PTFE specific enthalpy (gauge h=0 for PTFE at T0) as a function of temperature (at 1 bar). Thermal equilibrium for all C/F compounds (dashed); cases with CF2 (dotted) and C2F4 (solid) being the dominant decomposition products (see text for the meaning of T0 , Td , etc.).

This is illustrated for polytetrafluorethylene (PTFE, Teflon) in Fig. 3, where h(T) is shown for a mixture of compounds of C and F with ratio 1:2. Full equilibrium including all relevant species (atoms, molecules, solids, ions, ...) is represented by the dashed curve. Solid PTFE, however, can only be obtained from an equilibrium calculation (minimization of the Gibbs free enthalpy) if some specific compounds in the set of molar fractions are disregarded. For instance, solid carbon, gaseous CF4, C2F4, and others must be suppressed for CF2-dominant ablation (dotted curve). Then solid PTFE is stable and decomposes into CF2 at Td. A similar procedure is used for the C2F4-dominant case (solid curve), where gaseous C2F4 is the initial decomposition product. These curves describe partial thermodynamic equilibrium for a selected set of compounds. The neglect of specific compounds can be justified by chemical kinetics, since slowly forming products do not appear in the short dwell time in the Knudsen layer. Solid carbon formation, e.g., is too slow and thus negligible (the observed soot in breaker nozzles forms after arcing). According to the figure, the CF4 and C2F4 dominant cases equalize at a temperature (Te,1) larger than about 2000 K, and they equalize with the full equilibrium state for all species (Te,2) around 3500 K, being a typical value for the vapor temperature of PTFE ablation [9,10]. Note that the curves in Fig. 3 should not be confused with the state change across the Knudsen layer (which, by the way, encompasses a thin layer of the solid surface where polymer chain scission happens), in which even partial equilibrium is absent because the fragments are in highly excited nonequilibrium states. The curves in Fig. 3 are used for estimating h and Td of the ablation vapor outside the Knudsen layer.

The ablation boundary conditions must provide the mass current M , the normal velocity w, and the LTE temperature Tb at the boundary for a given heat power flux Qrad impinging the surface. Mass and energy balance read
M

Qrad

= w = M (h(Tb)-h(T0)+ w2)

(12) (13)

where we neglected additional possible power losses, and h of the vapor are functions of its (partial) pressure and temperature, and T0 is the initial (ambient) temperature of the solid. If ablation occurs into a dense gas and is not too strong, the Mach number w/wsound << 1, then w2 can be neglected in (13), and M = Qrad/(h(Tb)-h(T0)) (see also [3]). If the Mach number is not small, (12) can be used to eliminate w in (13). Hence, the problem is reduced to the determination of the vapor temperature Tb, which can be estimated in at least three different ways. Firstly, one can postulate on the grounds of (semi)empirical findings, a value, e.g., Tb = 3500 K for PTFE [9,10], giving a ablation enthalpy of h(Tb)-h(T0) 10-15 MJ/kg for 1-10 bar and T0 = 300 K. A second way refers to a Boltzmann transport equation treatment of the Knudsen layer [11,12]. However, the complexity of the decomposition kinetics makes an exact formulation impossible, in contrast to evaporation theory of simple molecular or atomic solids. Reference [12] made a step forward by taking internal degrees of freedom into account; however the details of polymer decomposition are too complex for a description with a single distribution function. A third approach [8] considers again the entropy production rate. In contrast to the previous section, it is not motivated by strict mathematical grounds [7] as for radiation modeling, but must be seen as an estimate [13] that takes the strongly chaotic mixing of many degrees

473

of freedom in the nonequilibrium layer into account. The underlying model-assumption is that equilibration during traversal of the ablated material through the Knudsen layer occurs as fast as possible. As fast is expressed by maximizing the (irreversible) entropy production rate in the Knudsen layer, given by

production maximization for a sub-set of ablation products. The models may handle the physical complexity, and provide an improved compromise between physical reliability and computational costs.

S (Tb) = M (s(Tb)-s(Td)) ,

(14)

which is a measure for the equilibration velocity. Here, s(Td) and s(Tb) are the specific entropies entering and leaving the Knudsen layer. Equilibration will be fast, because the responsible microscopic processes are very fast on a macroscopic time scale. But even if they were infinitely fast, equilibration will not occur immediately because of constraints. Hence, as possible means that the maximization is subject to constraints, which are given by the conservation laws (12) and (13) (at fixed p). It is straightforward to calculate the maximum of S as a function of p and Qrad, if the equations of states belong to an ideal gas and the vapor pressure curve is known. For real polymers, however, two problems appear. First, the maximization problem can be more complicated (for instance, more than a single local maximum may exist). The second and more severe problem refers to the uncertainty on the ablation products, and thus on which partial equilibrium enthalpy one should take. B. Illustrative Results To handle with this uncertainty, it is appropriate to consider different scenarios, e.g., those shown in Fig. 3. In Fig. 4, Tb is plotted for CF2-dominant (dotted) and C2F4-dominant (solid) ablation. Additionally, the values of Td, Te,1, and Te,2 are given. (In [8] we had used data for (CF2)n in full thermal equilibrium {dashed curve in Fig. 3}, which we know now is incorrect despite of the quantitative accordance with values in literature.) For clarity, we consider sufficiently small Qrad implying low Mach number and thus Tb-values independent of Qrad; this assumption is reasonable for typical cases [8]. One observes that our approach, despite its heuristic character, not only provides reasonable estimates (temperatures around 3500K) in accordance with expectations, but gives a range of uncertainty by scenario studies. It is now straightforward to estimate the ablation boundary conditions for other polymer materials. A simple example is polyethylene, where some obvious compounds like CH4 must be similarly disregarded in the decomposition kinetics; we find Tb values in the range of 1500 K. V. SUMMARY

Fig. 4. Temperatures for PTFE decomposition (Td), maximum entropy production (Tb), and equality of states (Te,1: CF2/C2F4, and Te,2: C2F4/all species) as functions of pressure (T0 =300 K).

ACKNOWLEDGEMENT The author thanks Frank Kassubek for very useful input. REFERENCES
[1] [2] [3] M. Kapetanovic, High Voltage Circuit Breakers, KEMA, ETF Sarajevo (2011). G. R. Jones and M. T. C. Fang The physics of high power arcs, Phys. Rep. 43, 1415-1465 (1980). R. Bini, B. Galletti, M. Schwinne, and T. Werder Schlpfer, CFD in Circuit Breaker Development, IEEE IECP1, pp ??. (2011). T. Christen, F. Kassubek, Minimum entropy production closure of the photo-hydrodynamic equations for radiative heat transfer, Journal of Quantitative Spectroscopy and Radiative Transfer, Vol. 110, 452-463 (2009). T. Christen, F. Kassubek, and R. Gati, Radiative Heat Transfer and effective Transport Coefficients, in Heat Transfer, edited by A. Belmiloudi, InTech, 101-126 (2011); (free online on intechweb.org). V. Aubrecht and M. Bartlowa, Net Emission Coefficients of Radiation in Air and SF6 Thermal Plasmas,Plasma Chem Plasma Process 29, 131147 (2009). J. M. Ziman, The general variational principle of transport theory, Can. J. Phys. 34, 1256-1273 (1956). T. Christen, A maximum entropy production model for teflon ablation by arc radiation, Journal of Physics D: Applied Physics Volume: 40, 5719-5722 (2007). M. T. C. Fang and D. B. Newland, DC nozzle arcs with mild wall ablation, J. Phys. D: Appl. Phys., 16, 793-810 (1983). C. B. Ruchti and L. Niemeyer, Ablation controlled arcs, IEEE Trans. Plasma. Science., PS-14, 423-434 (1986). M. Keidar, I. D. Boyd, and I. I. Beilis, On the model of Teflon Ablation in an ablation-controlled discharge, J. Phys. D: Appl. Phys. 34, 1675-1677 (2001). M. Zaghloul, On the vaporization of Teflon and heated compound-materials in ablation-controlled arcs, J. Appl. Phys., 95, 3339-3343 (2004). I. Ford and T-J. Lee, Entropy production and destruction in models of material evaporation J. Phys. D: Appl. Phys. 34, 413417 (2001).

[4]

[5]

[6]

[7] [8]

[9] [10] [11]

[12]

Entropy production based models of radiation and ablation for 3d CFD simulations of GCB were discussed. The essence of the radiation model is a closure of the moment equations by entropy production minimization. The ablation model is based on the replacement of the unspecified momentum balance equation by entropy

[13]

E-mail of authors: Thomas.christen@ch.abb.com

474

You might also like