issues with my first testcase

[yoanncheny]

Dear deepxde users,
I'm using deepxde for solving the BRCE equation which consists in 2 coupled PDE (1D & unsteady) with 2 unknown fields (h & R). By mimicking the tutorial examples I managed to build the PDE (the custom constants of the problem are not shown for clarity) :
`def pde(x, y):

h,r = y[:, 0:1], y[:, 1:2]
dh_x = dde.grad.jacobian(y, x, i=0, j=0)
dh_t = dde.grad.jacobian(y, x, i=0, j=1)
dr_x = dde.grad.jacobian(y, x, i=1, j=0)
dr_t = dde.grad.jacobian(y, x, i=1, j=1)    
dh_xx = dde.grad.hessian(y, x, component=0, i=0, j=0)
dr_xx = dde.grad.hessian(y, x, component=1, i=0, j=0)
source = -(gamma*dh_x+k*dh_xx)*r
eq_h = dh_t + source
eq_r = dr_t + vel*dr_x - diff * dr_xx - source 

return [eq_h,eq_r]`

the space-time domain :

geom = dde.geometry.Interval(0, L)
timedomain = dde.geometry.TimeDomain(0, Tf)
geomtime = dde.geometry.GeometryXTime(geom, timedomain)`

and the BCs and IC (Dirichlet on the right side of the domain L and Neumann and Robin on the left side for h & R )

`def boundary_l(x, on_boundary):
    return on_boundary and np.isclose(x[0], 0)

def boundary_r(x, on_boundary):
    return on_boundary and np.isclose(x[0], L)
def h_func(x):
    return h0-(h0/L)*x 
def r_func(x):
    return r0 

BC_h_right = dde.icbc.DirichletBC(
    geomtime, lambda X: 0, lambda _, on_boundary: boundary_r, component=0
)
BC_R_right = dde.icbc.DirichletBC(
    geomtime, lambda X: 0, lambda _, on_boundary: boundary_r, component=1
)
BC_h_left = dde.icbc.NeumannBC(
    geomtime, lambda X:0, lambda _, on_boundary: boundary_l, component=0
)
BC_R_left = dde.icbc.RobinBC(
    geomtime, lambda X,y: (vel*diff)*y , lambda _, on_boundary: boundary_l, component=1
)
initial_h = dde.icbc.IC(
    geomtime, h_func, lambda _, on_initial: on_initial, component=0
)
initial_r = dde.icbc.IC(
    geomtime, r_func, lambda _, on_initial: on_initial, component=1)

So far so good, but as I try to use Time PDE :

data = dde.data.TimePDE( geomtime, pde, [ BC_h_right, BC_R_right, BC_h_left, BC_R_left, initial_h, initial_r, ], num_domain=50000, num_boundary=5000, num_initial=2000, )

I get the following error message that I'm not able to solve :

`---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
~\AppData\Local\Temp\ipykernel_3884\1242027741.py in <module>
     12     num_domain=50000,
     13     num_boundary=5000,
---> 14     num_initial=10,
     15 )
     16 

~\anaconda3\envs\claw\lib\site-packages\deepxde\data\pde.py in __init__(self, geometryxtime, pde, ic_bcs, num_domain, num_boundary, num_initial, train_distribution, anchors, exclusions, solution, num_test, auxiliary_var_function)
    331             solution=solution,
    332             num_test=num_test,
--> 333             auxiliary_var_function=auxiliary_var_function,
    334         )
    335 

~\anaconda3\envs\claw\lib\site-packages\deepxde\data\pde.py in __init__(self, geometry, pde, bcs, num_domain, num_boundary, train_distribution, anchors, exclusions, solution, num_test, auxiliary_var_function)
    125         self.train_aux_vars, self.test_aux_vars = None, None
    126 
--> 127         self.train_next_batch()
    128         self.test()
    129 

~\anaconda3\envs\claw\lib\site-packages\deepxde\utils\internal.py in wrapper(self, *args, **kwargs)
     36             x = [getattr(self, a) for a in attr]
     37             if all(i is None for i in x):
---> 38                 return func(self, *args, **kwargs)
     39             return x if len(x) > 1 else x[0]
     40 

~\anaconda3\envs\claw\lib\site-packages\deepxde\data\pde.py in train_next_batch(self, batch_size)
    174     def train_next_batch(self, batch_size=None):
    175         self.train_x_all = self.train_points()
--> 176         self.bc_points()  # Generate self.num_bcs and self.train_x_bc
    177         if self.bcs and config.hvd is not None:
    178             num_bcs = np.array(self.num_bcs)

~\anaconda3\envs\claw\lib\site-packages\deepxde\utils\internal.py in wrapper(self, *args, **kwargs)
     36             x = [getattr(self, a) for a in attr]
     37             if all(i is None for i in x):
---> 38                 return func(self, *args, **kwargs)
     39             return x if len(x) > 1 else x[0]
     40 

~\anaconda3\envs\claw\lib\site-packages\deepxde\data\pde.py in bc_points(self)
    280     @run_if_all_none("train_x_bc")
    281     def bc_points(self):
--> 282         x_bcs = [bc.collocation_points(self.train_x_all) for bc in self.bcs]
    283         self.num_bcs = list(map(len, x_bcs))
    284         self.train_x_bc = (

~\anaconda3\envs\claw\lib\site-packages\deepxde\data\pde.py in <listcomp>(.0)
    280     @run_if_all_none("train_x_bc")
    281     def bc_points(self):
--> 282         x_bcs = [bc.collocation_points(self.train_x_all) for bc in self.bcs]
    283         self.num_bcs = list(map(len, x_bcs))
    284         self.train_x_bc = (

~\anaconda3\envs\claw\lib\site-packages\deepxde\icbc\boundary_conditions.py in collocation_points(self, X)
     50 
     51     def collocation_points(self, X):
---> 52         return self.filter(X)
     53 
     54     def normal_derivative(self, X, inputs, outputs, beg, end):

~\anaconda3\envs\claw\lib\site-packages\deepxde\icbc\boundary_conditions.py in filter(self, X)
     47 
     48     def filter(self, X):
---> 49         return X[self.on_boundary(X, self.geom.on_boundary(X))]
     50 
     51     def collocation_points(self, X):

IndexError: arrays used as indices must be of integer (or boolean) type

`

Could anyone help me ?
Kind regards,
Yoann

[yoanncheny]

Hi all,
I've made some progress debugging my code. By replacing

BC_h_right = dde.icbc.DirichletBC( geomtime, lambda X: 0, lambda _, on_boundary: boundary_r, component=0 )

with

BC_h_right = dde.icbc.DirichletBC( geomtime, lambda X: 0, lambda _, on_boundary: on_boundary, component=0 )

and doing the same for the other BC, the aforementioned error disappears. I recall the defintion of boundary_r (which seems to be responsible for the bug):

def boundary_r(x, on_boundary): return on_boundary and np.isclose(x[0], L)

Could anyone provide some help ?
Regards,
Yoann

[xuelanghanbao]

In my opinion, since you have defined the boundary function boundary_l and boundary_r, so you don't need to write the lambda expression. You can directly write it in the following form: BC_h_right = dde.icbc.DirichletBC( geomtime, lambda X: 0, boundary_r, component=0 )

[yoanncheny]

Thx for your answer. Now the code works !