Input Deck for Solver Parameters

This section details the solver parameters in the structured input deck (.par file), which is finally read by the main solver. In GUI mode, it is also the only purpose for the preprocessor to write out, and serves as the mid-file format for GUI resolve. It includes the path pointing to data, indicates the path to save results, and determines the modes and numerical methods for the solver. In other words, if the data is placed properly, this .par is the only input file by the solver.

Input and Range Control

Key	Description	Value(s)	Explaination
vtk_folder	Path to folder containing VTK input data	<string>
output_dir	Path to directory for saving results	<string>
auto_wall	Flag for automatic wall detection	0	Manually indicate computation bounds.
		(1)	Auto detect bounds from input mesh.
wall_xrange	X-direction wall bounds [xmin, xmax]	[<float>, <float>]	Dimensionalized coordinates bounds.
wall_yrange	Y-direction wall bounds [ymin, ymax]	[<float>, <float>]
wall_zrange	Z-direction wall bounds [zmin, zmax]	[<float>, <float>]
x	Number of Lagrangian seed points in X direction	<int>
y	Number of Lagrangian seed points in Y direction	<int>
z	Number of Lagrangian seed points in Z direction	<int>
T_range	Time range [T_start, T_end] in frames	[<float>, <float>]	T_end no larger than N_frames.
dt	Time step for particle advection in frames	<float>
physical_dt	Physical time step between two input frames	<float>	Perform physical advection.

Numerical Methods

Key	Description	Value(s)	Explaination
if_backward	Backward integration flag	0	Only compute forward advection (repelling LCS structures).
		(1)	Compute in both side (both repelling and attracting structures).
wall_treatment	Wall-treatment strategy	(clamp)	Out-of-bounds points take nearest velocity. Applicable to subdomains from larger ones.
		dirichlet	Simple and fast BC treatment, zeroing the boundary values. Introduces artifical shear.
		reflect	Reflect the particles back to the computational domain. Likely a non-shear wall BC.
		periodic	Wrap around to the opposite side. Applicable to fully-developed subdomains and periodic flows.
advc_method	Lagrangian advection scheme	(Euler)	Simple and fast Euler stepping with \(O(h^1)\) error.
		RK2	Runge–Kutta 2 with \(O(h^2)\) error.
		RK4	Runge–Kutta 4 with \(O(h^4)\) error.
		RK6	Runge–Kutta 6 with \(O(h^6)\) error.
interp_method	Interpolation method for velocity	trilinear	Simple and fast triliear interpolation method. Very fast on GPUs.
		(tricubic)	Catmull–Rom 3D convolution kernel for tricubic interpolation.
		hermite
		WENO	Weighted Essentially Non-Oscillatory (WENO) scheme with shockwave capture capability.
		tricubicFL	I.P. The high-performance 3D tricubic interpolation method by [Lekien2005].
grad_order	Method of gradient discretization	(2)	2nd-order central difference.
		4	I.P. 4th-order central difference.
		6	I.P. 6th-order central difference.
		100	I.P. Global Fast Fourier Transform (FFT) with finite order. Compute extremely intensive!
eigen_method	Max eigenvalue of C-G tensor	(eigmax_sym3)	Closed-form cubic eigenvalue solver for symmetric 3×3 matrices (noniterative solver).
		jacobi	I.P. Jacobi eigenvalue algorithm (iterative).

Dynamic LCS Window

Key	Description	Value(s)	Explaination
dyn_mode	Dynamic LCS mode selector	(0)	Treat this input as a steady whole, generating still results. See Steady LCS.
		1	Dynamic LCS computation based on sliding window method. See Unsteady LCS.
win_size	Window size for dynamic LCS	<float>	The window size in frames for each batch of computation. Available under dyn_mode=1.
win_step	Window step for dynamic LCS	<float>	Stepping in frames. Better smaller than win_size/2 for capturing dynamic patterns.

Utilizations

Key	Description	Value(s)	Explaination
if_visual	Glace visualization enabler	0 / 1	Take a pop-up glance on FTLE coutours after computation for check with a rudimentary appearance.