ORNL Seminar 19.05.11 Reactor Point Kinetics-- Then and Now Barry D. Ganapol Fellow
description
Transcript of ORNL Seminar 19.05.11 Reactor Point Kinetics-- Then and Now Barry D. Ganapol Fellow
ORNL Seminar19.05.11
Reactor Point Kinetics--Then and Now Barry D. Ganapol
Fellow Advanced Institute of Studies Unibo
and DIENCA Visiting Professor
andUniversity of Arizona
UTK
v Errors and missing information in papers
v Unsubstantiated claims of accuracy, simplicity, usefulness and elegance
v Lack of benchmarks and benchmarking strategyv The “simple” algorithm is missingv Extreme accuracy has always been achievablev No ultimate PKE algorithm currently exists
Some thoughts concerning PKE algorithms--
Sketch of Point Kinetics Equations
N(t)
dN(t)
N(t)-
N(t)+ +S(t)
Separability:
Adjoint weighting:m
i = 1,…,m
A Survey of Past Solutions to the PKEs
Nuclear Reactor Kineticsby G.R. Keepin, 1965
1
1
, 1,...,
n
i n i n
n
t m
n i i nit
tt t t ti
i i nt
N t N t dt t N t C t C tl
C t e C t dt N t e i ml
+ Laplace transform and inversion + Requires extensive tables of poles and residuals + RTS code advanced at the time + Too difficult to use routinely
+ Integral form (RTS Code 1960)
A New Solution of the Point Kinetics EquationsJ. A. W. da Nóbrega
NSE 46, 366-375 (1971)
v Consider constant reactivity insertion and constant source
d ttdt
yAy S
0
0
0
1
10 ...m
m
NN
N
y
10t e e At AtA I Sy y
1 2
11
2
....
0 0 .... 0 ....
.... 0
0 .... 0
m
mm
t
A
1
...m
N tC tt
C t
y
v Require inverse of A which is argued to be too computationally expensive (at the time) - Advocates Padé approximant, e.g., P(2,0)
3 32 / 6f At he At A
12 22 1 / 2 , , ,i if At h h fcn
A A
Note: All eigenvalues are not necessary but still require extreme eigenvalues
v At best 5x10-5 relative error
Unnecessarily complicated for outcome
Solution of the Reactor Kinetics Equations by Analytical Continuation
John Vigil NSE 29, 292-401 (1967)
Taylor Series
Recurrence
+ Time step control
+ Continuous Analytical Continuation
Time Discretization
+ A method ahead of its time
On the Numerical Solution of the Point KineticsEquations by Generalized Runge-Kutta Methods
J. SanchezNSE 103, 94-99 (1989)
0, 0y t f y t y y
*1
1
1* *
1
1*
1
, 1,...,
s
n n k kk
k
n k n kj jj
k
n kj jj
y y c f
I hf y f hf y f
hf y f k s
All coefficients are specified
Method Underperforms
t GRK Claimed
Exact Exact
0.003 1 2.20985 2.2098 2.20984045698 10 8.01891 8.0192 8.01919997323 20 2.82948e+1 2.8351e+1 2.82973997810e+1 0.0055 0.1 5.21 5.21 5.21002839001 2 4.3022e+1 4.3025e+1 4.30251435157e+1 10 1.388e+5 1.3886e+5 1.38860225602e+5 0.007 0.01 4.50885 4.5088 4.50885848635 0.5 5.3445e+3 5.3459e+3 5.34588761204e+3 2 2.05697e+11 2.0591e+11 2.05915601782e+11 0.008 0.01 6.20276 6.2029 6.20285357509 0.1 1.4101e+3 1.4104e+3 1.41042180359e+3 1 6.1486e+23 6.1634e+23 6.16333374991e+23
L = 2e-05s
A new integral method for solving the pointreactor neutron kinetics equations
Li, Chen, Luo, Zhu, ANE 36, 427-432 (2009)
v Start from integral equation and assume
21
0
10
, 1,...,i
kjh
n jj
kh
i n i n j ijj
N t a e
C t e C t a G i m
t BBF Claimed Exact Exact =-0.003
0.2 4.809750E-01 4.809743E-01 4.80973210584E-01 0.4 4.652904E-01 4.652903E-01 4.65289326117E-01 0.6 4.519648E-01 4.519650E-01 4.51963975793E-01 0.8 4.402738E-01 4.402732E-01 4.40272277652E-01 1.0 4.297825E-01 4.297830E-01 4.29782046265E-01
= 0.007 0.2 1.597867E+02 1.597257E+02 1.59725769863E+02 0.4 1.665854E+03 1.667386E+03 1.66728769255E+03 0.6 1.709542E+04 1.713190E+04 1.71319342103E+04 0.8 1.752925E+05 1.758905E+05 1.75890975931E+05 1.0 1.797286E+06 1.805726E+06 1.80573163423E+06
L = 2e-05sMethod performs poorly
Aboanber Methods:PWS: an efficient code system for solving
space-independent nuclear reactor dynamicsA.E. Aboanber*, Y.M. Hamada
Annals of Nuclear Energy 29 (2002) 2159–2172
2.000000000E+00 1.338200050E+00 4.000000000E+00 2.228441897E+00 6.000000000E+00 5.582052449E+00 8.000000000E+00 4.278629573E+01 9.000000000E+00 4.875200217E+02 1.000000000E+01 4.511636239E+05 1.100000000E+01 1.792213607E+16
Ramp
Claimed “exact” solution is not so
Solution of the point kinetics equations in the presence
of Newtonian temperature feedback by Pad´eapproximations via the analytical inversion method
A E Aboanber and A A NahlaJ. Phys. A: Math. Gen. 35 (2002) 9609–9627
Same method as da Nóbrega and Sanchez
2.000000000E+00 1.338200050E+00 4.000000000E+00 2.228441897E+00 6.000000000E+00 5.582052449E+00 8.000000000E+00 4.278629573E+01 9.000000000E+00 4.875200217E+02 1.000000000E+01 4.511636239E+05 1.100000000E+01 1.792213607E+16
Inaccurate at large time
A Resolution of the Stiffness Problem ofReactor KineticsY. Chao, A. Attard
NSE 90, 40-46 (1985)
0 0( ) , ( ) ( )
t t
dt w t dt u t
i iN t e C t f t e
Define:
Choose u and w to confine most variation to N
t SCM Claimed
Exact Exact
0.003 1 2.2254 2.2098 2.20984045698 10 8.0324 8.0192 8.01919997323 20 2.8351e+1 2.8297e+1 2.82973997810e+1 0.0055 0.1 5.2057 5.2100 5.21002839001 2 4.3024e+1 4.3025e+1 4.30251435157e+1 10 1.3875e+5 1.3886e+5 1.38860225602e+5 0.007 0.01 4.5001 4.5088 4.50885848635 0.5 5.3530e+3 5.3459e+3 5.34588761204e+3 2 2.0627e+11 2.0591e+11 2.05915601782e+11 0.008 0.01 6.2046 6.2029 6.20285357509 0.1 1.4089e+3 1.4104e+3 1.41042180359e+3 1 6.1574e+23 6.1634e+23 6.16333374991e+23
Confusing mathematical physics modeling
An analytical solution of the point kinetics equations with time-variable reactivity
by the decomposition methodClaudio Z. Petersen , Sandra Dulla, Marco T.M.B.
Vilhena, Piero RavettoProgress in Nuclear Energy xxx (2011)
Adomian Decomposition Method
Inappropriate claims of accuracy, utility and simplicity
t 0.003 0.007 0.008
1.0000E-05 1.02940792213E+00 1.07000000355E+00 1.08040134076E+00 1.0000E-03 1.73656448024E+00 8.00355370443E+00 1.47531198687E+01 1.0000E-01 1.80476704702E+00 1.50688886725E+04 2.89485876592E+44 1.0000E+00 2.21520297121E+00 2.85050383437E+25 3.25148973415+436 1.0000E+01 8.05231886873E+00 1.66961653500+238 1.32953281684+872* *t = 2 (QP)
L = 10-6s
A simple reliable, robust algorithm to solve the PKEs is lacking.
+ Must think numerically and use - new computational architectures - robust numerical methods - experimental numerical methods
+ Must abandon the outmoded ideas of time step control and the “minimum time step competition”.
….hence….
A Survey of New Solutions to the PKEs + GPCA + TS
1
,
, 1,...,
m
l ll
l ll l
dN t t NN t C t
dt
dC tN t C t l m
dt
m GPCA G(?)Piecewise Constant Approximation to the Solution of Point Kinetics Eqns (PKEs)
Recall:
,d t
t N t tdt y
A y
1 2
11
2
, ....
0 0, .... 0 ....
.... 0
0 .... 0
m
mm
R t N t
t N t
A
,,
t N tR t N t
v First consider constant reactivity insertion (without source)
d ttdt
yAy
,0t e Aty y 1
1
1
0 ...m
m
y
Diagonalize A : A = UWU-1
1 0Det
I A+ Eigenvalues from
+ Eigenvectors form U from
0 k UI A 1k k
+ U-1 = VT from transpose0 T T
k VI A
+ W = diag{k; k=1,…,G}T VUWA
Attributable to da Nobrega, Sanchez, Allen, Aboanber
1
1
2,..., 1/ , 1,..., 11
; , 1,..., 1
llk k
k l
lk
l mk m
l k m
u
U u
2
1
/11
ml l
k kl k l
1
1
22
1
1/ 1
1
; , 1,..., 1
k llk
k k l
ml l
k kl k l
lk l k m
v
vV
In-hour Equation:
0Tt e VWtUy y
+ Exact solution for step insertion
HQR Algorithm for k
- Algorithm Summary
Explicit eigenvector representation
Note: All done through linear algebra
1
11
1 , i
i
t
i i ii i t
dt t t t tt t
11
0j ji i
iT T
i i i i j jj
e e
v vW tW tu uy y y
v Now consider prescribed reactivity insertion
- Note: must introduce a time step and solve for eigenvalues for each time interval
Efficient numerical solution of the point kineticsequations in nuclear reactor dynamics
M. Kinard and E. AllenANE 31 1039-1051 (2004)
An Implicit Method for Solving the Lumped Parameter Reactor-Kinetics Equations by
Repeated ExtrapolationM. Izumi and T NodaNSE 41 299-303 (1970)
Combined R-K with repeated Richardsonextrapolation to improve the FD/RK scheme
Can this concept be generalized ?
Another article method ahead of its time
v Convergence acceleration + True solution based on the limit
1
0lim 0j
iT
j jj
hhi
t e
vWuy y
+ Form a sequence of solutions
1
; , , 0,1,....2
0j ni
T ii n j j n n
j
h tt h h ne
vWuy y
;limi i nnt t h
y y
+ Apply a convergence accelerator to to find a new sequence such that
0;
lim i n i
i i nn
t tt t h
y yy y
n ity
n ity is found from the asymptotic behavior (in n) of original sequence
Some accelerators are: Romberg, Aitkin Wynn-epsilon (W-e), Euler transformation
+GPCA Ganapolized Piecewise Constant Reactivity Approximation
tj-1 tj
Sequentially halve interval Build a sequence of solutions over all gridsAccelerate convergence via Romberg or W-eBegin each interval with converged IC
- Goal is extreme accuracy
lr (n = 2lr)
2 4 6 8 10 12 14 16
Rel
ativ
e Er
ror
1e-13
1e-12
1e-11
1e-10
1e-9
1e-8
1e-7
1e-6
1e-5
1e-4
1e-3
1e-2
1e-1
1e+0
1e+1
No AccelerationRombergW-e
Ramp 0.1/sFor all edits
Approach to Prompt Jump
Time (s)0 2x100 4x100 6x100 8x100 10x100 12x100
Neu
tron
Den
sity
0
2
4
6
8
10
12
14
= 10-7,10-8,….,10-19
Approach to Prompt Jump
Time (s)-24-23-22-21-20-19-18-17-16-15-14-13-12-11-10-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3
Neu
tron
Den
sity
0
2
4
6
8
10
12
14
=10-19
= 10-7
$0.50 Step Insertion in a fast reactor
2.0000E+00 1.3382000E+00 4.0000E+00 2.2284419E+00 6.0000E+00 5.5820524E+00 8.0000E+00 4.2786295E+01
Claudio Z. Petersen , Sandra Dulla, Marco T.M.B. Vilhena, Piero Ravetto
Progress in Nuclear Energy xxx (2011)
GPCA
Ramp
1.0000E+00 1.1239405E+00 2.0000E+00 1.1688896E+00 3.0000E+00 1.0744847E+00 4.0000E+00 9.5382929E-01 5.0000E+00 9.0735349E-01 1.0000E+01 9.8468032E-01
0.00073sint t 19652011
Here is robust for ramp considered earlier 2.0000E+00 1.338200050E+00 4.0000E+00 2.228441897E+00 6.0000E+00 5.582052449E+00 8.0000E+00 4.278629573E+01 9.0000E+00 4.875200217E+02 1.0000E+01 4.511636239E+05 1.1000E+01 1.792213607E+16
GPCA Correct to 9-places in comparisonwith FD (Ganapol/NSE Letter)
v Test by manufactured solution
1 0
1 ii
tmt tt
i ii
dN tt
N t dt
e dt e N tN t
+ Solve for reactivity
+ Specify N(t)+ Input imply t N t
1 1 tN t f e
1
1 11
i i
i
t
t tm
i tti
i
t eN t
e f e
fN t e e
Assume exponential to power level
Implies
Manufactured Solution f = 2
X Data
0 1 2 3 4 5 6
Den
sity
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Time(s)0 1 2 3 4 5 6
Rea
ctiv
ity
0.000
0.001
0.002
0.003
0.004
0.005
Time(s)0 1 2 3 4 5 6
Num
ber o
f Cor
rect
Dig
its
8
9
10
11
12
13
14
15
Error Measure forall 3 cases
= 10-10
m Taylor series (TS) solution to PKEs (J. Vigil/1967) + GPCA - requires discretization solution and In-hour - not analytical - iteration to include non-linear reactivity
+ TS solution most natural solution and gives an analytical solution
,
, , 10
,
k jk
i i k j jk
k j
N t nc t c t t
t
+ Form TS in interval [tj-1,tj]
+ Naturally generate the following recurrence:
1, , , ,1
, 1, , , ,
11
1
m
k j k j i i k ji
ii k j k j i i k j
k n c
k c n c
, 0, ,
1
, ,0
k j j k j
k
k l j l jl
n
n
+ Numerical implementation - Must proceed with caution TS slowly converging and therefore sensitive to round off (from “swell”) - Use Continuous Analytical Continuation (CAC)
, 1
0
0, 1
k
k j jk
j j
N t n t t
n N t
Choose interval [tj-1,tj] to limit number of terms in TS to KAccelerate convergence of partial sums via W-eat original and added time edits if necessary
tj-1 tj
- Form sequence for N(tj) by refining grids
then accelerate grid endpoint with W-e
Called “effort based time step control”
1.0000E+00 1.130832298566E+00 2.0000E+00 1.338200050049E+00 3.0000E+00 1.668749984373E+00 4.0000E+00 2.228441896810E+00 5.0000E+00 3.277251197891E+00 6.0000E+00 5.582052448674E+00 7.0000E+00 1.216798285040E+01 8.0000E+00 4.278629573112E+01 9.0000E+00 4.875200217231E+02 1.0000E+01 4.511636239090E+05 1.1000E+01 1.792213607343E+16
1.130832298566E+00 1.338200050049E+00 1.668749984373E+00 2.228441896810E+00 3.277251197891E+00 5.582052448673E+00 1.216798285040E+01 4.278629573111E+01 4.875200217231E+02 4.511636239090E+05 1.792213607342E+16
GPCA TS
Comparison for 0.1$/s Ramp
1
0
1 1
0, 1
1, 1
, 1,
, 2
j
t
t
j jt
j j
j j
k j k j
t at B dt N t
t t a t t B dt N t
t
a BN t
Bn k
+ Demonstration with feedback - Adiabatic temperature feedback (Keepin)
Adiabatic with Doppler
Time(s)0 1 2 3 4 5 6
Neu
tron
Den
sity
10-1
100
101
102
103
104
105
106
107
108
109
1010
1011
1012
a = 0.003, 0.005, 0.01, 0.05,0.1
0 1 2 3 4 5 610-1
100
101
102
103
104
105
106
107
108
109
1010
1011
0 1 2 3 4 5 6
Der
ivat
ive
-1e+11
-8e+10
-6e+10
-4e+10
-2e+10
0
2e+10
4e+10
6e+10
8e+10
1e+11
1
, m
l ll
dN t t NN t C t
dt
Determination of time a first peak
2.91058215.114160E+09TS
Most accurate calculation to date
My Message:+ Two highly accurate solutions to PKE presented
+ All based on previous work
+ Achievable in 1970
+ Methods far outperform any existing algorithm for standard problems of prescribed and non-linear reactivities
+ Simple and elegant and sets the standard
+ Nearly completes the numerical solution to PKEs