Asking about PointSetBC and NeumannBC

[DanHoleKimson]

Hello

I am a university student who wishes to solve Maxwell's equations using deepxde. I'm encountering an issue with the following problem:
u_xx + u_yy = 0 in Ω
n ∇u = g in ∂Ω
u_0 = 0

I have set up PointSetBC, but I couldn't achieve
U(0,0)=0, and I'm facing anomalies in values around (-1, 0). I would appreciate it if you could help me identify any mistakes I might have made.

Here's my code:

import deepxde as dde
import numpy as np
import torch
import matplotlib.pyplot as plt

R = 1
num_dense_layers = 5
num_dense_nodes = 150
activation = "tanh"

geom = dde.geometry.Disk([0, 0], R)

center = np.array([0, 0])
def func(x):
    return np.zeros([len(x), 1])

def pde(x, y):
    du_xx = dde.grad.hessian(y, x, i=0, j=0)
    du_yy = dde.grad.hessian(y, x, i=1, j=1)
    return du_xx + du_yy


def boundary_bottom(x, on_boundary):
    return on_boundary and (np.cos(9*np.pi/16) <= x[0] <= np.cos(7*np.pi/16)) and x[1] <= -np.sin(7*np.pi/16)

def boundary_top(x, on_boundary):
    return on_boundary and (np.cos(9*np.pi/16) <= x[0] <= np.cos(7*np.pi/16)) and x[1] >= np.sin(7*np.pi/16)

def boundary_other1(x, on_boundary):
    return on_boundary and ((np.cos(9*np.pi/16) >= x[0] >= -1) or (np.cos(7*np.pi/16) <= x[0] <= 1))

bc_c = dde.icbc.PointSetBC(center, func(center))
bc_r = dde.icbc.NeumannBC(geom, lambda x : -1, boundary_bottom)
bc_t = dde.icbc.NeumannBC(geom, lambda x : 1, boundary_top)
bc_o1 = dde.icbc.NeumannBC(geom, lambda x : 0, boundary_other1)

data = dde.data.PDE(
    geom, pde, [bc_c, bc_t, bc_r, bc_o1], num_domain=3000, num_boundary=150,
)

layer_size = [2] + [num_dense_nodes] * num_dense_layers + [1]
net = dde.nn.pytorch.fnn.FNN(layer_size, activation,"Glorot uniform")

model = dde.Model(data, net)
model.compile("adam", lr=1e-3) 
losshistory, train_state = model.train(iterations=15000)
dde.saveplot(losshistory,train_state,issave=False,isplot=True)

Thank you.