1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409 | // *****************************************************************************
/*!
\file src/PDE/PrefIndicator.cpp
\copyright 2012-2015 J. Bakosi,
2016-2018 Los Alamos National Security, LLC.,
2019-2021 Triad National Security, LLC.
All rights reserved. See the LICENSE file for details.
\brief Adaptive indicators for p-adaptive discontiunous Galerkin methods
\details This file contains functions that provide adaptive indicator
function calculations for marking the number of degree of freedom of each
element.
*/
// *****************************************************************************
#include "PrefIndicator.hpp"
#include "Tags.hpp"
#include "Vector.hpp"
#include "Integrate/Basis.hpp"
#include "Integrate/Quadrature.hpp"
namespace inciter {
void spectral_decay( std::size_t nmat,
std::size_t nunk,
const std::vector< int >& esuel,
const tk::Fields& unk,
std::size_t ndof,
std::size_t ndofmax,
tk::real tolref,
std::vector< std::size_t >& ndofel )
// *****************************************************************************
//! Evaluate the spectral-decay indicator and mark the ndof for each element
//! \param[in] nmat Number of materials in this PDE system
//! \param[in] nunk Number of unknowns
//! \param[in] esuel Elements surrounding elements
//! \param[in] unk Array of unknowns
//! \param[in] ndof Number of degrees of freedom in the solution
//! \param[in] ndofmax Max number of degrees of freedom for p-refinement
//! \param[in] tolref Tolerance for p-refinement
//! \param[in,out] ndofel Vector of local number of degrees of freedome
//! \details The spectral decay indicator, implemented in this functiopn,
//! calculates the difference between the projections of the numerical
//! solutions on finite element space of order p and p-1.
//! \see F. Naddei, et. al., "A comparison of refinement indicators for the
//! p-adaptive simulation of steady and unsteady flows with discontinuous
//! Galerkin methods" at https://doi.org/10.1016/j.jcp.2018.09.045, and G.
//! Gassner, et al., "Explicit discontinuous Galerkin schemes with adaptation
//! in space and time"
// *****************************************************************************
{
const auto ncomp = unk.nprop() / ndof;
// The array storing the adaptive indicator for each elements
std::vector< tk::real > Ind(nunk, 0);
for (std::size_t e=0; e<esuel.size()/4; ++e) {
if(ndofel[e] > 1) {
if(nmat == 1)
Ind[e] =
evalDiscIndicator_CompFlow(e, ncomp, ndof, ndofel[e], unk);
else if(nmat > 1)
Ind[e] =
evalDiscIndicator_MultiMat(e, nmat, ncomp, ndof, ndofel[e], unk);
}
}
// As for spectral-decay indicator, rho_p - rho_(p-1) actually is the leading
// term of discretization error for the numerical solution of p-1. Therefore,
// this function represents the qualitative behavior of the discretization
// error. If the value is less than epsL which means the discretization error
// is already a really small number, then the element should be de-refined. On
// the other hand, if the value is larger than epsH which means the
// discretization error is relatively large, then it should be refined.
// Note: Spectral-decay indicator is a measurement of the continuity of the
// numerical solution inside this element. So when this indicator appears
// to be relatively large, there might be a shock inside this element and a
// derefinement or h-refinement should be applied. This condition will be
// implemented later.
// As for the discretiazation-error based indicator, like spectral-decay
// indicator, the choices for epsH and epsL are based on the data from
// numerical experiments. Empirically, we found that when the epsH belongs
// to [-4, -8] and epsL belongs to [-6, -14], decent results could be
// observed. And then a linear projection is performed to map epsL and espH
// to the range of [0, 1] so that it could be controlled by tolref.
auto epsH = std::pow(10, -4 - tolref * 4.0);
auto epsL = std::pow(10, -6 - tolref * 8.0);
// The epsL_p2 as well as the 'else if' code below are kept, since sometimes
// just using a common epsL for both DGP2 and DGP1 might not produce good ndof
// distributions. Further testing required by manupulating these 3 thresholds.
//auto epsL_p2 = std::pow(10, -7 - tolref * 8.0);
// Marke the ndof according to the adaptive indicator
for (std::size_t e=0; e<esuel.size()/4; ++e)
{
if(Ind[e] < epsL) // Derefinement
{
if(ndofel[e] == 4)
ndofel[e] = 1;
else if(ndofel[e] == 10)
ndofel[e] = 4;
}
//else if (Ind[e] < epsL_p2 && ndofel[e] == 10) {
// ndofel[e] = 4;
//}
else if(Ind[e] > epsH) // Refinement
{
if(ndofel[e] == 4 && ndofmax > 4)
ndofel[e] = 10;
}
}
}
void non_conformity( std::size_t nunk,
std::size_t Nbfac,
const std::vector< std::size_t >& inpoel,
const tk::UnsMesh::Coords& coord,
const std::vector< int >& esuel,
const std::vector< int >& esuf,
const std::vector< std::size_t >& inpofa,
const tk::Fields& unk,
std::size_t ndof,
std::size_t ndofmax,
std::vector< std::size_t >& ndofel )
// *****************************************************************************
//! Evaluate the non-conformity indicator and mark the ndof for each element
//! \param[in] nunk Number of unknowns
//! \param[in] Nbfac Number of internal faces
//! \param[in] inpoel Element-node connectivity
//! \param[in] coord Array of nodal coordinates
//! \param[in] esuel Elements surrounding elements
//! \param[in] esuf Elements surrounding faces
//! \param[in] inpofa Face-node connectivity
//! \param[in] unk Array of unknowns
//! \param[in] ndof Number of degrees of freedom in the solution
//! \param[in] ndofmax Max number of degrees of freedom for p-refinement
//! \param[in,out] ndofel Vector of local number of degrees of freedome
//! \details The non-conformity indicator, this function implements, evaluates
//! the jump in the numerical solutions as a measure of the numerical error.
//! \see F. Naddei, et. al., "A comparison of refinement indicators for the
//! p-adaptive simulation of steady and unsteady flows with discontinuous
//! Galerkin methods at https://doi.org/10.1016/j.jcp.2018.09.045.
//! \warning This indicator can only be applied in serial, i.e., single CPU, for
//! now because the solution communication happens before eval_ndof() in DG,
//! which will lead to incorrect evaluation of the numerical solution at the
//! neighboring cells.
// *****************************************************************************
{
const auto ncomp = unk.nprop() / ndof;<--- Variable 'ncomp' is assigned a value that is never used.
const auto& cx = coord[0];
const auto& cy = coord[1];
const auto& cz = coord[2];
// The array storing the adaptive indicator for each elements
std::vector< tk::real > Ind(nunk, 0);
// compute error indicator for each face
for (auto f=Nbfac; f<esuf.size()/2; ++f)
{
Assert( esuf[2*f] > -1 && esuf[2*f+1] > -1, "Interior element detected "
"as -1" );
std::size_t el = static_cast< std::size_t >(esuf[2*f]);
std::size_t er = static_cast< std::size_t >(esuf[2*f+1]);
auto ng_l = tk::NGfa(ndofel[el]);
auto ng_r = tk::NGfa(ndofel[er]);
// When the number of gauss points for the left and right element are
// different, choose the larger ng
auto ng = std::max( ng_l, ng_r );
// arrays for quadrature points
std::array< std::vector< tk::real >, 2 > coordgp;
std::vector< tk::real > wgp;
coordgp[0].resize( ng );
coordgp[1].resize( ng );
wgp.resize( ng );
// get quadrature point weights and coordinates for triangle
tk::GaussQuadratureTri( ng, coordgp, wgp );
// Extract the element coordinates
std::array< std::array< tk::real, 3>, 4 > coordel_l {{
{{ cx[ inpoel[4*el ] ], cy[ inpoel[4*el ] ], cz[ inpoel[4*el ] ] }},
{{ cx[ inpoel[4*el+1] ], cy[ inpoel[4*el+1] ], cz[ inpoel[4*el+1] ] }},
{{ cx[ inpoel[4*el+2] ], cy[ inpoel[4*el+2] ], cz[ inpoel[4*el+2] ] }},
{{ cx[ inpoel[4*el+3] ], cy[ inpoel[4*el+3] ], cz[ inpoel[4*el+3] ] }} }};
std::array< std::array< tk::real, 3>, 4 > coordel_r {{
{{ cx[ inpoel[4*er ] ], cy[ inpoel[4*er ] ], cz[ inpoel[4*er ] ] }},
{{ cx[ inpoel[4*er+1] ], cy[ inpoel[4*er+1] ], cz[ inpoel[4*er+1] ] }},
{{ cx[ inpoel[4*er+2] ], cy[ inpoel[4*er+2] ], cz[ inpoel[4*er+2] ] }},
{{ cx[ inpoel[4*er+3] ], cy[ inpoel[4*er+3] ], cz[ inpoel[4*er+3] ] }} }};
// Compute the determinant of Jacobian matrix
auto detT_l =
tk::Jacobian( coordel_l[0], coordel_l[1], coordel_l[2], coordel_l[3] );
auto detT_r =
tk::Jacobian( coordel_r[0], coordel_r[1], coordel_r[2], coordel_r[3] );
// Extract the face coordinates
std::array< std::array< tk::real, 3>, 3 > coordfa {{
{{ cx[ inpofa[3*f ] ], cy[ inpofa[3*f ] ], cz[ inpofa[3*f ] ] }},
{{ cx[ inpofa[3*f+1] ], cy[ inpofa[3*f+1] ], cz[ inpofa[3*f+1] ] }},
{{ cx[ inpofa[3*f+2] ], cy[ inpofa[3*f+2] ], cz[ inpofa[3*f+2] ] }} }};
// Gaussian quadrature
for (std::size_t igp=0; igp<ng; ++igp)
{
// Compute the coordinates of quadrature point at physical domain
auto gp = tk::eval_gp( igp, coordfa, coordgp );
//Compute the basis functions
auto B_l = tk::eval_basis( ndofel[el],
tk::Jacobian( coordel_l[0], gp, coordel_l[2], coordel_l[3] ) / detT_l,
tk::Jacobian( coordel_l[0], coordel_l[1], gp, coordel_l[3] ) / detT_l,
tk::Jacobian( coordel_l[0], coordel_l[1], coordel_l[2], gp ) / detT_l );
auto B_r = tk::eval_basis( ndofel[er],
tk::Jacobian( coordel_r[0], gp, coordel_r[2], coordel_r[3] ) / detT_r,
tk::Jacobian( coordel_r[0], coordel_r[1], gp, coordel_r[3] ) / detT_r,
tk::Jacobian( coordel_r[0], coordel_r[1], coordel_r[2], gp ) / detT_r );
std::array< std::vector< tk::real >, 2 > state;
state[0] = tk::eval_state( ncomp, ndof, ndofel[el], el, unk, B_l );
state[1] = tk::eval_state( ncomp, ndof, ndofel[er], er, unk, B_r );<--- Variable 'state[1]' is assigned a value that is never used.
Assert( unk[0].size() == ncomp, "Size mismatch" );
Assert( unk[1].size() == ncomp, "Size mismatch" );
auto rhoL = state[0][0];
auto rhoR = state[1][0];
auto ind = fabs( rhoL - rhoR ) / 2.0 * ( rhoL + rhoR );
Ind[el] = std::max( ind, Ind[el] );
Ind[er] = std::max( ind, Ind[er] );
}
}
// By assuming a smooth solution, we use the non-conformity indicator to
// represent the error for the numerical solution qualitatively. If the value
// is less than epsL which means the error is already a really small number,
// then the element should be de-refined. On the other hand, if the value is
// larger than epsH which means the error is relatively large, then it should
// be refined.
// Marke the ndof according to the adaptive indicator
for (std::size_t e=0; e<esuel.size()/4; ++e)
{
if(Ind[e] < 1e-4) // Derefinement
{
if(ndofel[e] == 10)
ndofel[e] = 4;
else if(ndofel[e] == 4)
ndofel[e] = 1;
}
else if(Ind[e] > 1e-3) // Refinement
{
if(ndofel[e] == 4 && ndofmax > 4)
ndofel[e] = 10;
else if(ndofel[e] == 1)
ndofel[e] = 4;
}
}
}
tk::real evalDiscIndicator_CompFlow( std::size_t e,
ncomp_t ncomp,
const std::size_t ndof,
const std::size_t ndofel,
const tk::Fields& unk )
// *****************************************************************************
//! Evaluate the spectral decay indicator
//! \param[in] e Index for the tetrahedron element
//! \param[in] ncomp Number of scalar components in this PDE system
//! \param[in] ndof Number of degrees of freedom in the solution
//! \param[in] ndofel Local number of degrees of freedom
//! \param[in] unk Array of unknowns
//! \return The value of spectral indicator for the element
//! \detail The spectral indicator evaluates the density differences between
//! the numerical solutions at different polynomial space
// *****************************************************************************
{
auto ng = tk::NGvol(ndofel);
// arrays for quadrature points
std::array< std::vector< tk::real >, 3 > coordgp;
std::vector< tk::real > wgp( ng );
coordgp[0].resize( ng );
coordgp[1].resize( ng );
coordgp[2].resize( ng );
tk::GaussQuadratureTet( ng, coordgp, wgp );
tk::real dU(0.0), U(0.0), Ind(0.0);
// Gaussian quadrature
for (std::size_t igp=0; igp<ng; ++igp)
{
// Compute the basis function
auto B = tk::eval_basis( ndofel, coordgp[0][igp], coordgp[1][igp],
coordgp[2][igp] );
auto state = tk::eval_state( ncomp, ndof, ndofel, e, unk, B );
U += wgp[igp] * state[0] * state[0];
if(ndofel > 4)
{
auto dU_p2 = unk(e, 4) * B[4]
+ unk(e, 5) * B[5]
+ unk(e, 6) * B[6]
+ unk(e, 7) * B[7]
+ unk(e, 8) * B[8]
+ unk(e, 9) * B[9];
dU += wgp[igp] * dU_p2 * dU_p2;
}
else
{
auto dU_p1 = unk(e, 1) * B[1]
+ unk(e, 2) * B[2]
+ unk(e, 3) * B[3];
dU += wgp[igp] * dU_p1 * dU_p1;
}
}
Ind = dU / U;
return Ind;
}
tk::real evalDiscIndicator_MultiMat( std::size_t e,
std::size_t nmat,
ncomp_t ncomp,
const std::size_t ndof,
const std::size_t ndofel,
const tk::Fields& unk )
// *****************************************************************************
//! Evaluate the spectral decay indicator
//! \param[in] e Index for the tetrahedron element
//! \param[in] nmat Number of materials in this PDE system
//! \param[in] ncomp Number of scalar components in this PDE system
//! \param[in] ndof Number of degrees of freedom in the solution
//! \param[in] ndofel Local number of degrees of freedom
//! \param[in] unk Array of unknowns
//! \return The value of spectral indicator for the element
//! \detail The spectral indicator evaluates the bulk density differences
//! between the numerical solutions at different polynomial space
// *****************************************************************************
{
auto ng = tk::NGvol(ndof);
// arrays for quadrature points
std::array< std::vector< tk::real >, 3 > coordgp;
std::vector< tk::real > wgp( ng );
coordgp[0].resize( ng );
coordgp[1].resize( ng );
coordgp[2].resize( ng );
tk::GaussQuadratureTet( ng, coordgp, wgp );
tk::real dU(0.0), U(0.0), Ind(0.0);
// Gaussian quadrature
for (std::size_t igp=0; igp<ng; ++igp)
{
// Compute the basis function
auto B = tk::eval_basis( ndof, coordgp[0][igp], coordgp[1][igp],
coordgp[2][igp] );
auto state = tk::eval_state( ncomp, ndof, ndofel, e, unk, B );
tk::real denom(0.0), numer(0.0);
for(std::size_t k = 0; k < nmat; k++) {
denom += state[densityIdx(nmat, k)];
if(ndofel > 4) {
numer += ( unk(e, densityDofIdx(nmat, k, ndof, 4)) * B[4]
+ unk(e, densityDofIdx(nmat, k, ndof, 5)) * B[5]
+ unk(e, densityDofIdx(nmat, k, ndof, 6)) * B[6]
+ unk(e, densityDofIdx(nmat, k, ndof, 7)) * B[7]
+ unk(e, densityDofIdx(nmat, k, ndof, 8)) * B[8]
+ unk(e, densityDofIdx(nmat, k, ndof, 9)) * B[9] );
} else {
numer += ( unk(e, densityDofIdx(nmat, k, ndof, 1)) * B[1]
+ unk(e, densityDofIdx(nmat, k, ndof, 2)) * B[2]
+ unk(e, densityDofIdx(nmat, k, ndof, 3)) * B[3] );
}
}
dU += wgp[igp] * numer * numer;
U += wgp[igp] * denom * denom;
}
Ind = dU / U;
return Ind;
}
}
// inciter::
|