-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathaorta_FEA_C3D8_SRI_R1.py
381 lines (374 loc) · 16.4 KB
/
aorta_FEA_C3D8_SRI_R1.py
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
import sys
sys.path.append("./c3d8")
sys.path.append("./mesh")
import numpy as np
from IPython import display
import matplotlib.pyplot as plt
import torch
from torch.linalg import det
from AortaFEModel_C3D8_SRI import AortaFEModel, cal_potential_energy, cal_element_orientation
from FEModel_C3D8_SRI_fiber import cal_attribute_on_node
from FEModel_C3D8 import cal_F_tensor_8i, cal_F_tensor_1i
from von_mises_stress import cal_von_mises_stress
from PolyhedronMesh import PolyhedronMesh
from aorta_mesh import get_solid_mesh_cfg
import time
#%%
matA="200, 0, 1, 0.3333, 0, 1e5"
matB="50, 1000, 10, 0.3333, 0, 1e5"
matC="50, 1000, 10, 0.1, 60, 1e5"
mat_sd="1e5, 0, 1, 0.3333, 0, 1e5"
all_mat=torch.load('./data/125mat.pt')['mat_str']
mat95=all_mat[95]
mat10=all_mat[10]
mat24=all_mat[24]
mat64=all_mat[64]
matMean=torch.load('./data/125mat.pt')['mean_mat_str']
#%%
folder_data='./data/'
folder_result='./result/forward/'
#%%
import argparse
parser = argparse.ArgumentParser(description='Input Parameters:')
parser.add_argument('--cuda', default=0, type=int)
parser.add_argument('--dtype', default="float64", type=str)
parser.add_argument('--shell_template', default=folder_data+'bav17_AortaModel_P0_best.pt', type=str)
parser.add_argument('--mesh_p0', default=folder_data+'p0_47_solid', type=str)
parser.add_argument('--mesh_px_pred', default=folder_result+'p0_47_solid_matMean_p18_pred_refine', type=str)
parser.add_argument('--mesh_px_init', default=folder_result+'p0_47_solid_matMean_p18_pred', type=str)
#parser.add_argument('--mesh_px_init', default='', type=str)
parser.add_argument('--mat', default=matMean, type=str)
parser.add_argument('--pressure', default=18, type=float)
parser.add_argument('--max_iter1', default=1, type=int)
parser.add_argument('--max_iter2', default=1000, type=int)
parser.add_argument('--use_stiffness', default=True, type=bool)
parser.add_argument('--reform_stiffness_interval', default=20, type=int)
parser.add_argument('--warm_up_T0', default=10, type=int)
parser.add_argument('--ratio_Interval', default=10, type=int)
parser.add_argument('--TPE_ratio', default=0.01, type=float)
parser.add_argument('--loss1', default=0.01, type=float)
parser.add_argument('--Rmax', default=0.005, type=float)
parser.add_argument('--save_by_vtk', default='True', type=str)#True: save *.vtk file in addition to *.pt file
parser.add_argument('--plot', default='False', type=str)
arg = parser.parse_args()
arg.save_by_vtk = True if arg.save_by_vtk == 'True' else False
arg.plot = True if arg.plot == 'True' else False
print(arg)
#%%
device=torch.device("cuda:"+str(arg.cuda))
#device=torch.device("cpu")
if arg.dtype == "float64":
dtype=torch.float64
elif arg.dtype == "float32":
dtype=torch.float32
else:
raise ValueError("unkown dtype:"+arg.dtype)
#%%
n_layers=1
if '4layer' in arg.mesh_p0:
n_layers=4
(boundary0, boundary1, Element_surface_pressure, Element_surface_free)=get_solid_mesh_cfg(arg.shell_template, n_layers)
#%%
Mesh_X=PolyhedronMesh()
Mesh_X.load_from_torch(arg.mesh_p0+".pt")
Node_X=Mesh_X.node.to(dtype).to(device)
Element=Mesh_X.element.to(device)
#%%
from Mat_GOH_SRI import cal_1pk_stress, cal_cauchy_stress
#from InvivoMat_GOH_SRI import cal_1pk_stress, cal_cauchy_stress
Mat=[float(m) for m in arg.mat.split(",")]
Mat[4]=np.pi*(Mat[4]/180)
Mat=torch.tensor([Mat], dtype=dtype, device=device)
aorta_model=AortaFEModel(None, Element, Node_X, boundary0, boundary1, Element_surface_pressure,
Mat, cal_1pk_stress, cal_cauchy_stress, dtype, device, mode='inflation')
aorta_model.element_orientation_on_deformed_mesh=False
pressure=arg.pressure
#%%
def loss1a_function(force_int, force_ext):
loss=((force_int-force_ext)**2).sum(dim=1).sqrt().mean()
return loss
#%%
def loss1b_function(force_int, force_ext):
loss=((force_int-force_ext)**2).sum()
return loss
#%%
if len(arg.mesh_px_init) > 0:
Mesh_x_init=PolyhedronMesh()
Mesh_x_init.load_from_torch(arg.mesh_px_init+".pt")
Node_x_init=Mesh_x_init.node.to(dtype).to(device)
u_field_init=Node_x_init-Node_X
u_field_init=u_field_init*aorta_model.state['mask']
print("initilize u_field with u_field_init")
aorta_model.state['u_field'].data=u_field_init[aorta_model.state['free_node']].to(dtype).to(device)
out=aorta_model.cal_energy_and_force(pressure)
TPE1=out['TPE1']; TPE2=out['TPE2']; SE=out['SE']
force_int=out['force_int']; force_ext=out['force_ext']
force_int_of_element=out['force_int_of_element']
F=out['F']; u_field=out['u_field']
loss1=loss1a_function(force_int, force_ext)
loss1=loss1.item()
force_avg=(force_int_of_element**2).sum(dim=2).sqrt().mean()
force_res=((force_int-force_ext)**2).sum(dim=1).sqrt()
R=force_res/(force_avg+1e-10)
Rmean=R.mean().item()
Rmax=R.max().item()
J=det(F)
print("init: pressure", pressure)
print("init: TPE", float(TPE1), "loss1", loss1, "Rmax", Rmax, "max|J-1|", float((J-1).abs().max()))
sum_force_res=force_res.sum().item()
print("init: sum(force_res)", force_res.sum().item(), "1/sum(force_res)", 1/sum_force_res)
#1/sum_force_res should be similar to t_default for the first step
if loss1 < arg.loss1 and Rmax < arg.Rmax:
opt_cond=True
arg.max_iter1=0
print("mesh_px_init is optimal, set arg.max_iter1=0, set opt_cond=True")
#%%
del Mesh_x_init, Node_x_init, u_field_init
del out, TPE1, TPE2, SE, force_int, force_ext, J, F, u_field, R, force_res, force_avg, force_int_of_element
#%%
loss1_list=[]
loss2_list=[]
Rmax_list=[]
TPE_list=[]
time_list=[]
flag_list=[]
t0=time.time()
#%%
from FE_lbfgs import LBFGS
optimizer = LBFGS([aorta_model.state['u_field']], lr=1, line_search_fn="backtracking",
tolerance_grad =1e-10, tolerance_change=1e-20, history_size =100, max_iter=1)
optimizer.set_backtracking(c=0.5, verbose=False)
t_listA=[1, 0.5, 0.1, 0.05, 0.01]
t_listB=[0.5, 0.3, 0.1, 0.05, 0.01]
#%%
def reform_stiffness(pressure_t, iter2):
ta=time.time()
Output=aorta_model.cal_energy_and_force(pressure_t, return_stiffness="sparse")
H=Output['H']
H=H.astype("float64")
optimizer.reset_state()
optimizer.set_H0(H)
tb=time.time()
print("reform stiffness done:", tb-ta, "iter2", iter2)
return tb-ta
#%%
max_iter1=arg.max_iter1
max_iter2=arg.max_iter2
Ugood=aorta_model.state['u_field'].clone().detach()
closure_opt={"output":"TPE", "loss1_fn":loss1a_function}
#%%
for iter1 in range(0, max_iter1):
optimizer.reset_state()
pressure_iter=((iter1+1)/max_iter1)*pressure
print("pressure_iter", pressure_iter)
error_flag=False
flag_use_stiffness=False
optimizer.reset_state()
optimizer.set_backtracking(t_list=t_listA, t_default=0.5, t_default_init="auto")
#t_default_init for the first step must be very small to prevent nan
for iter2 in range(0, max_iter2):
#------------------------------------------------
def closure(output=closure_opt["output"], loss1_fn=closure_opt["loss1_fn"]):
out=aorta_model.cal_energy_and_force(pressure_iter)
TPE1=out['TPE1']; TPE2=out['TPE2']; SE=out['SE']
force_int=out['force_int']; force_ext=out['force_ext']
F=out['F']; u_field=out['u_field']
loss1=loss1_fn(force_int, force_ext)
force_res=force_int-force_ext
loss2=cal_potential_energy(force_res, u_field)
if output == "TPE":
loss=loss2
elif output == "loss1":
loss=loss1
elif output == "TPE+loss1":
loss=loss2+loss1
else:
loss=loss2
if loss.requires_grad==True:
optimizer.zero_grad()
loss.backward()
if output == "TPE":
return TPE1
elif output == "loss1":
return loss1
elif output == "TPE+loss1":
return TPE+loss1
else:
loss1=float(loss1)
loss2=float(loss2)
TPE1=float(TPE1)
F=F.detach()
force_int=force_int.detach()
force_ext=force_ext.detach()
force_int_of_element=out['force_int_of_element'].detach()
return loss1, loss2, TPE1, F, force_int, force_ext, force_int_of_element
#------------------------------------------------
opt_cond=optimizer.step(closure)
flag_linesearch=optimizer.get_linesearch_flag()
t_linesearch=optimizer.get_linesearch_t()
#------------------------------------------------
loss1, loss2, TPE, F, force_int, force_ext, force_int_of_element=torch.no_grad()(closure)(output='all')
J=det(F)
#------------------------------------------------
#check error
error_flag_counter=0
#TPE is nan or inf?
if (np.isnan(TPE) == True or np.isinf(TPE) == True
or np.isnan(loss1) == True or np.isinf(loss1) == True
or np.isnan(loss2) == True or np.isinf(loss2) == True):
opt_cond=False
error_flag=True
error_flag_counter+=1
print(iter1, iter2, "error: TPE", TPE, "loss1", loss1, "loss2", loss2, "max|J-1|", float((J-1).abs().max()))
print("flag", flag_linesearch, "t_linesearch", t_linesearch)
aorta_model.state['u_field'].data=Ugood.clone().detach()
optimizer.reset_state()
optimizer.set_backtracking(t_list=t_listA, t_default=0.5, t_default_init="auto")
flag_use_stiffness=False
if iter2==0:
print("error at iter2 = 0, break")
break
continue
#It is possible that TPE > 0, not as bad as TPE=nan or +/- inf
if TPE > 0:
opt_cond=False
error_flag=True
error_flag_counter+=1
#print(iter1, iter2, "error: TPE is", TPE, "loss1 is", loss1, ", max|J-1| =", float((J-1).abs().max()))
#print("flag", flag_linesearch, "t_linesearch", t_linesearch)
if flag_use_stiffness == True:
aorta_model.state['u_field'].data=Ugood.clone().detach()
optimizer.reset_state()
optimizer.set_backtracking(t_list=t_listA, t_default=0.5, t_default_init=0.5)
flag_use_stiffness=False
continue
#some elements may deform too much
J_error_counter=((J-1).abs()>0.5).sum().item()
if J_error_counter > 0:
opt_cond=False
error_flag=True
error_flag_counter+=1
if flag_use_stiffness == True:
aorta_model.state['u_field'].data=Ugood.clone().detach()
optimizer.reset_state()
optimizer.set_backtracking(t_list=t_listA, t_default=0.5, t_default_init=0.5)
flag_use_stiffness=False
if iter2 > arg.warm_up_T0:
print(iter1, iter2, "error: sum(abs(J-1)>0.5) =", J_error_counter)
print("flag", flag_linesearch, "t_linesearch", t_linesearch)
continue
#set error_flag
if error_flag_counter == 0:
error_flag=False
#------------------------------------------------
if iter2 >= 1 and error_flag==False and flag_use_stiffness == False:
optimizer.set_backtracking(t_list=t_listA, t_default=0.5, t_default_init=0.5)
#------------------------------------------------
flag_list.append(flag_linesearch)
t1=time.time()
time_list.append(t1-t0)
TPE_list.append(TPE)
loss1_list.append(loss1)
loss2_list.append(loss2)
force_avg=(force_int_of_element**2).sum(dim=2).sqrt().mean()
force_res=((force_int-force_ext)**2).sum(dim=1).sqrt()
R=force_res/(force_avg+1e-10)
Rmean=R.mean().item()
Rmax=R.max().item()
Rmax_list.append(Rmax)
#-------------------------------------------------------
if error_flag == False and len(TPE_list) > 1:
if TPE < min(TPE_list[:-1]) or loss1 < min(loss1_list[:-1]) or Rmax < min(Rmax_list[:-1]):
Ugood=aorta_model.state['u_field'].clone().detach()
#-------------------------------------------------------
T=arg.ratio_Interval
T0=max(arg.warm_up_T0, T)
if len(TPE_list) < T0:
TPE_ratio=1
loss1_ratio=1
if len(TPE_list) > 2:
id_T=max(-T, -len(TPE_list))
TPE_ratio=abs(TPE_list[-1]-TPE_list[id_T])/(1e-10+abs(TPE_list[-1]))
loss1_ratio=abs(loss1_list[-1]-loss1_list[id_T])/(1e-10+abs(loss1_list[0]))
else:
TPE_ratio=abs(TPE_list[-1]-TPE_list[-T])/(1e-10+abs(TPE_list[-1]))
loss1_ratio=abs(loss1_list[-1]-loss1_list[-T])/(1e-10+abs(loss1_list[0]))
if error_flag == False and flag_use_stiffness == False and arg.use_stiffness==True:
if TPE_ratio < arg.TPE_ratio: #loss1_ratio is not good for input close to optimal (loss1 very small)
flag_use_stiffness=True
print("iter1", iter1, "iter2", iter2, "pressure_iter", pressure_iter, "time", t1-t0)
print("flag", flag_linesearch, "t_linesearch", t_linesearch, "TPE_ratio", TPE_ratio, "loss1_ratio", loss1_ratio)
print('Rmax', Rmax, "loss1", loss1, "loss2", loss2, "TPE", TPE, "max|J-1|", float((J-1).abs().max()))
print('set flag_use_stiffness to True')
#-------------------------------------------------------
if arg.use_stiffness == True and flag_use_stiffness == True and error_flag == False:
if iter2%arg.reform_stiffness_interval == 0:
reform_stiffness(pressure_iter, iter2)
#loss nan if the solution is not close to the true optimal one
optimizer.set_backtracking(t_list=t_listB, t_default=0.5, t_default_init=0.5)
#------------------------------------------
opt_cond=False
#check convergance
if loss1 < arg.loss1 and Rmax < arg.Rmax:
opt_cond=True
#------------------------------------------
flag_print=False
if iter2==0:
flag_print=True
if flag_use_stiffness==False and iter2%T == 0:
flag_print=True
if arg.use_stiffness==True and flag_use_stiffness==True and (iter2+1)%arg.reform_stiffness_interval==0:
flag_print=True
if flag_print==True or opt_cond == True:
print("iter1", iter1, "iter2", iter2, "pressure_iter", pressure_iter, "time", t1-t0)
print("flag", flag_linesearch, "t_linesearch", t_linesearch, "TPE_ratio", TPE_ratio, "loss1_ratio", loss1_ratio)
print('Rmax', Rmax, "loss1", loss1, "loss2", loss2, "TPE", TPE, "max|J-1|", float((J-1).abs().max()))
if arg.plot == True:
display.clear_output(wait=False)
fig, ax = plt.subplots()
ax.plot(np.array(loss1_list)/max(loss1_list), 'r')
ax.plot(loss1_list, 'm')
if len(TPE_list) > 10:
negTPE=-np.array(TPE_list)
Vmax=np.max(negTPE[(np.isinf(negTPE)==False)&(negTPE>0)])
ax.plot(negTPE/Vmax, 'b')
ax.plot(-np.array(flag_list), 'g.')
ax.set_ylim(0, 1)
ax.grid(True)
display.display(fig)
plt.close(fig)
#-------------------------------------
if opt_cond == True:
print("opt_cond is True, break iter2 =", iter2)
Ugood=aorta_model.state['u_field'].clone().detach()
error_flag=False
break
#%%
aorta_model.state['u_field']=Ugood
u_field=aorta_model.get_u_field()
Node_x=Node_X+u_field
S=aorta_model.cal_stress('cauchy', create_graph=False, return_W=False)
S_element=S.mean(dim=1)
VM_element=cal_von_mises_stress(S_element)
S_node=cal_attribute_on_node(Node_x.shape[0], Element, S_element)
VM_node=cal_von_mises_stress(S_node)
Mesh_x=PolyhedronMesh()
Mesh_x.element=Element.detach().cpu()
Mesh_x.node=Node_x.detach().cpu()
Mesh_x.element_data['S']=S_element.view(-1,9)
Mesh_x.element_data['VM']=VM_element.view(-1,1)
Mesh_x.node_data['S']=S_node.view(-1,9)
Mesh_x.node_data['VM']=VM_node.view(-1,1)
Mesh_x.mesh_data['arg']=arg
Mesh_x.mesh_data['TPE']=TPE_list
Mesh_x.mesh_data['loss1']=loss1_list
Mesh_x.mesh_data['Rmax']=Rmax
Mesh_x.mesh_data['time']=time_list
Mesh_x.mesh_data['opt_cond']=opt_cond
mesh_px_pred=arg.mesh_px_pred
if opt_cond == False:
mesh_px_pred=arg.mesh_px_pred+'_error'
Mesh_x.save_by_torch(mesh_px_pred+".pt")
if arg.save_by_vtk == True:
Mesh_x.save_by_vtk(mesh_px_pred+".vtk")
print("saved", mesh_px_pred)