-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathpredict.py
233 lines (184 loc) · 6.42 KB
/
predict.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
# %%
import deepxde as dde
from MyCuboid import *
from RandomCuboidHeatSourceMultiLayer import *
import pyvista as pv
from utils import *
import deepxde
import deepxde as dde
from deepxde.nn import NN
import torch
import numpy as np
from DeepONetV2 import DeepONet_V2
from RandomCuboidHeatSourceMultiLayer import RandomCuboidHeatSourceMultiLayer
from CustomPDEOperator import CustomPDEOperator
# 参考量
L_ref = 0.001 # mm
T_ref = 273.15 # K
k_ref = 398 #
q_heat = 1e11 * L_ref**2 / (k_ref * T_ref)
DD = 131/k_ref
# 定义PDE方程
def pde_heat(x, y,var):
"""
x: 输入坐标 (N, 3) -> (x, y, z)
y: 网络输出 (N, 1) -> T(x, y, z)
f: 额外的源项或可选参数, 例如热源 f(x,y,z)
返回值: PDE 残差, 在稳态情况下要求残差=0
"""
# 假设导热系数 D 是常数，也可以根据需要改成函数形式
# D = 131
# dx，dy，dz是1，1，0.1 还是 3.2，3.2，0.1
D = y[1][0]
y = y[0][:, 0, :]
Qheat = torch.zeros_like(D)
Qheat[D == DD] = q_heat
# 计算二阶偏导数 (拉普拉斯算子)
# T_xx, T_yy, T_zz 分别是对 x, y, z 的二阶偏导
T_xx = dde.grad.hessian(y, x, i=0, j=0)
T_yy = dde.grad.hessian(y, x, i=1, j=1)
T_zz = dde.grad.hessian(y, x, i=2, j=2)
# 稳态热传导方程: -D * (T_xx + T_yy + T_zz) = v
# 若 v=0, 表示无热源
return D * (T_xx + T_yy + T_zz) + Qheat
# %%
import deepxde as dde
import numpy as np
from CustomBC import CustomNeumannBC, CustomRobinBC
# 定义几何体
geo_jiban = MyCuboid([-5, -5, -1.02], [5, 5, -0.02])
geo_chiplayer = MyCuboid([-1.6, -1.6, -0.02], [1.6, 1.6, 0.36])
geo_heatspead = MyCuboid([-5, -5, 0.36], [5, 5, 2.36])
geo_heatsink = MyCuboid([-7.5, -7.5, 2.36], [7.5, 7.5, 6.36])
# 组合几何体
geo = dde.geometry.CSGUnion(dde.geometry.CSGUnion(geo_jiban, geo_chiplayer), geo_heatspead)
geo = dde.geometry.CSGUnion(geo, geo_heatsink)
# 侧壁 定义边界条件
def boundary(x, on_boundary):
if not on_boundary:
return False
# 排除 z = -1.02/1000 和 z = 6.36/1000 的点
z_min = -1.02
z_max = 6.36
return not (dde.utils.isclose(x[2], z_min) or dde.utils.isclose(x[2], z_max))
def boundary_top(x, on_boundary):
return on_boundary and dde.utils.isclose(x[2], 6.36)
def boundary_bot(x, on_boundary):
return on_boundary and dde.utils.isclose(x[2], -1.02)
def robin_func_top(x, y):
# y = y[:, 0, :]
D = 398 / k_ref
H = 5000 * L_ref / k_ref
# A = 15 * 15 / 1000000
t_ab = (273.15 + 25) / T_ref - 1
return -H / D * (y - t_ab)
def robin_func_bot(x, y):
# y = y[:, 0, :]
D = 130 / k_ref
H = 10 * L_ref / k_ref
# A = 10 * 10 / 1000000
t_ab = (273.15 + 25) / T_ref - 1
return -H / D * (y - t_ab)
bc_top = CustomRobinBC(geom=geo, func=robin_func_top, on_boundary=boundary_top)
bc_bot = CustomRobinBC(geom=geo, func=robin_func_bot, on_boundary=boundary_bot)
# 侧壁法向梯度为零的条件
bc_cebi = CustomNeumannBC(geo, lambda x: 0, boundary)
data = dde.data.PDE(
geo,
pde=pde_heat,
bcs=[bc_cebi, bc_top, bc_bot],
num_domain=1000,
num_boundary=1000,
train_distribution="uniform",
anchors=None
)
# %%
from MyCuboid import MyCuboid
space = RandomCuboidHeatSourceMultiLayer(
# Lx=3.2/1000,
# Ly=3.2/1000,
Lz=0.1,
dx=1,
dy=1,
dz=0.01,
n_sources_per_layer=3,
intensity=q_heat, # 设置热源强度
max_attempts=200
)
# 定义离散热源空间所需几何体
geo_heatspace1 = MyCuboid([-1.6, -1.6, 0.09], [1.6, 1.6, 0.1])
geo_heatspace2 = MyCuboid([-1.6, -1.6, 0.21], [1.6, 1.6, 0.22])
geo_heatspace3 = MyCuboid([-1.6, -1.6, 0.33], [1.6, 1.6, 0.34])
# 划分离散热源空间的网格点
eval_pts1 = geo_heatspace1.uniform_points_xyz(nx=32, ny=32, nz=3, boundary=True)
eval_pts2 = geo_heatspace2.uniform_points_xyz(nx=32, ny=32, nz=3, boundary=True)
eval_pts3 = geo_heatspace3.uniform_points_xyz(nx=32, ny=32, nz=3, boundary=True)
eval_pts = np.vstack((eval_pts1, eval_pts2, eval_pts3))
dataOperator = CustomPDEOperator(data, space, eval_pts, num_function=1, num_test=1)
# %%
net = DeepONet_V2(
[9216, 128, 128, 128],
[3, 128, 128, 128],
"tanh",
"Glorot normal",
)
model = dde.Model(dataOperator, net)
model.compile("adam", lr=0.0005)
# %%
new_model = model.restore(save_path="model_best.ckpt-19800.pt")
# %%
# 网格化空间
geo_jiban = MyCuboid([-5, -5, -1.02], [5, 5, -0.02])
geo_chiplayer = MyCuboid([-1.6, -1.6, -0.02], [1.6, 1.6, 0.36])
geo_heatspead = MyCuboid([-5, -5, 0.36], [5, 5, 2.36])
geo_heatsink = MyCuboid([-7.5, -7.5, 2.36], [7.5, 7.5, 6.36])
x_jiban = geo_jiban.uniform_points_xyz(nx=100,ny=100,nz=10)
x_chiplayer = geo_chiplayer.uniform_points_xyz(nx=128,ny=128,nz=38)
x_heatspead = geo_heatspead.uniform_points_xyz(nx=100,ny=100,nz=20)
x_heatsink = geo_heatsink.uniform_points_xyz(nx=150,ny=150,nz=40)
x_pre_all = np.vstack((x_jiban, x_chiplayer, x_heatspead, x_heatsink))
# %%
# print(type(x),x.shape)
# %%
#随机热源
space = RandomCuboidHeatSourceMultiLayer(
# Lx=3.2/1000,
# Ly=3.2/1000,
Lz=0.1,
dx=1,
dy=1,
dz=0.01,
n_sources_per_layer=3,
intensity=q_heat, # 设置热源强度
max_attempts=200
)
# 定义离散热源空间所需几何体
geo_heatspace1 = MyCuboid([-1.6, -1.6, 0.09], [1.6, 1.6, 0.1])
geo_heatspace2 = MyCuboid([-1.6, -1.6, 0.21], [1.6, 1.6, 0.22])
geo_heatspace3 = MyCuboid([-1.6, -1.6, 0.33], [1.6, 1.6, 0.34])
# 划分离散热源空间的网格点
eval_pts1 = geo_heatspace1.uniform_points_xyz(nx=32, ny=32, nz=3, boundary=True)
eval_pts2 = geo_heatspace2.uniform_points_xyz(nx=32, ny=32, nz=3, boundary=True)
eval_pts3 = geo_heatspace3.uniform_points_xyz(nx=32, ny=32, nz=3, boundary=True)
eval_pts = np.vstack((eval_pts1, eval_pts2, eval_pts3))
func_feats_pre = space.random(1)
np.savetxt("func_feats_pre.csv", func_feats_pre, delimiter=",")
v = space.eval_batch(func_feats_pre, eval_pts)
x_list = [
('x_jiban', x_jiban),
('x_chiplayer', x_chiplayer),
('x_heatspead', x_heatspead),
('x_heatsink', x_heatsink),
('x_pre_all', x_pre_all)
]
for name, x in x_list:
try:
D_factor = getDFactorV3(x, func_feats_pre)
y = model.predict((v, x, D_factor))
temperature_data = y[0][:, 0, 0]
temperature_data = (temperature_data + 1) * T_ref
grid = pv.PolyData(x)
grid.point_data['Temperature'] = temperature_data
grid.save(f"{name}.vtk")
except Exception as e:
print(f"Error processing {name}: {e}")