What is signed distance function (SDF) doing in integral continuity planes?

[prakhar_sharma]

In the Scalar Transport: 2D Advection Diffusion example, I am confused with how the integral continuity planes have been implemented.

# integral continuity
integral_continuity = IntegralBoundaryConstraint(
nodes=nodes,
geometry=integral_line,
outvar={"normal_dot_vel": 1},
batch_size=cfg.batch_size.num_integral_continuity,
integral_batch_size=cfg.batch_size.integral_continuity,
lambda_weighting={"normal_dot_vel": 0.1},
criteria=geo.sdf > 0,
param_ranges=x_pos_range,
)
domain.add_constraint(integral_continuity, "integral_continuity")

Here, I understand that batch_size denotes the number of integral continuity planes and integral_batch_size denotes the number of points to be sampled in each integral continuity plane. These things can be changed in the config file.

I also understand that a very low priority has been given to the targeted mass flow rate with a lambda_weighting of 0.1.

However, I am confused with what criteria=geo.sdf > 0 is doing.

Here geo = channel section - 3 rectangular fins

Is there a way to plot the sdf values, as an example plt.scatter(x,y,c=geo.sdf) to see where it goes negative.

I pulled out the definition of Line as integral continuity planes are Line object, but could not get any strong conclusion.

class Line(ConstructiveSolidGeometry):
    """
    2D Line parallel to y-axis

    Parameters
    ==========
    point_1 : tuple with 2 ints or floats
      lower bound point of line segment
    point_2 : tuple with 2 ints or floats
      upper bound point of line segment
    normal : int or float
      normal direction of line (+1 or -1)
    """

    def __init__(self, point_1, point_2, normal):
        assert point_1[0] == point_2[0], "Points must have same x-coordinate"

        # make sympy symbols to use
        l = Symbol(csg_curve_naming(0))
        x = Symbol("x")

        # curves for each side
        ranges = {l: (0, 1)}
        dist_y = point_2[1] - point_1[1]
        line_1 = Curve(
            functions={
                "x": point_1[0],
                "y": point_1[1] + l * dist_y,
                "normal_x": 1e-10 + normal,  # TODO rm 1e-10
                "normal_y": 0,
            },
            ranges=ranges,
            area=dist_y,
        )
        curves = [line_1]

        # calculate SDF
        sdf = normal * (point_1[0] - x)

        # initialize Line
        super(Line, self).__init__(curves, sdf)

[ohennigh]

Hello, so one tip is to visualize the integral constraint in paraview. When you run this it will make a file in the network directory call constraints. This will have vtp files for a single batch of all the constraints. You can load these into Paraview to see them. The reason we have that criteria=geo.sdf > 0 is we don’t want to integrate points that are inside those 3 fins. When you plot them you will see that if the integral line intersects the fins then it will not sample points there. This allows the integral to properly be computed.