|
MFC
Exascale flow solver
|
|
MFC
Exascale flow solver
|
Contains module m_ibm. More...
Go to the source code of this file.
Modules | |
| module | m_ibm |
| Ghost-node immersed boundary method: locates ghost/image points, computes interpolation coefficients, and corrects the flow state. | |
Functions/Subroutines | |
| impure subroutine, public | m_ibm::s_initialize_ibm_module () |
| Allocates memory for the variables in the IBM module. | |
| impure subroutine, public | m_ibm::s_ibm_setup () |
| Initializes the values of various IBM variables, such as ghost points and image points. | |
| subroutine, private | m_ibm::s_compute_ghost_point_pressure (gp, gp_patch_id, rho, pres_ip, pres_gp) |
| Pressure correction for a moving IB, accounting for the acceleration of the boundary surface. Clamped both ways: the linearization it comes from holds only while the correction is order one, and an unbounded one drives the ghost state to vacuum. | |
| subroutine, private | m_ibm::s_compute_ghost_point_velocity (gp, gp_patch_id, radial_vector, vel_ip, pres_ip, vel_gp) |
| subroutine, public | m_ibm::s_ibm_correct_state (q_cons_vf, q_prim_vf, pb_in, mv_in) |
| Update the conservative variables at the ghost points. | |
| impure subroutine, public | m_ibm::s_report_ibm_surface |
| One line at the end of a run if the reacting surface ever failed to solve. Silence means every ghost point on every step reached the surface-balance tolerance, i.e. the surface chemistry was actually applied everywhere it was asked for. | |
| impure subroutine, private | m_ibm::s_compute_image_points () |
| Compute the image points for each ghost point. | |
| subroutine, private | m_ibm::s_find_num_ghost_points (num_gps_out) |
| Count the number of ghost points for memory allocation. | |
| subroutine, private | m_ibm::s_find_ghost_points () |
| Locate all ghost points in the domain. | |
| subroutine, private | m_ibm::s_compute_interpolation_coeffs () |
| Compute the interpolation coefficients for image points. | |
| subroutine, private | m_ibm::s_interpolate_image_point (q_prim_vf, gp, alpha_rho_ip, alpha_ip, pres_ip, vel_ip, c_ip, r_ip, v_ip, pb_ip, mv_ip, nmom_ip, pb_in, mv_in, presb_ip, massv_ip, ys_ip) |
| Interpolate primitive variables to a ghost point's image point using bilinear or trilinear interpolation. | |
| impure subroutine | m_ibm::s_update_mib (num_ibs) |
| Resets the current indexes of immersed boundaries and replaces them after updating the position of each moving immersed boundary. | |
| pure real(wp) function | m_ibm::f_log_cosh (x) |
| log(cosh(x)) without overflow for large |x| | |
| subroutine | m_ibm::s_prescribed_kinematics (i, t) |
| Prescribed kinematics for patch i at time t: hinged flapping (kin_model = 1) or the Eldredge pitch ramp (kin_model = 2). Flapping rolls phi about the lab x axis through the hinge and pitches theta about the body spanwise (y) axis through it; the ramp holds phi at zero and drives theta alone. Both compose as R = Rx(phi) Ry(theta) (MFC's angle convention) and set the angles, centroid, velocity and lab-frame angular velocity; nothing is integrated, so restarts are exact. | |
| subroutine | m_ibm::s_compute_ib_forces (q_prim_vf, fluid_pp) |
| Compute pressure and viscous forces and torques on immersed bodies via volume integration. | |
| subroutine | m_ibm::s_compute_centroid_offset (gid, ib_marker) |
| Computes the center of mass for IB patch types where we are unable to determine their center of mass analytically. These patches include things like NACA airfoils and STL models Move a moving airfoil/STL patch's centroid to its centre of mass, keeping the difference in centroid_offset. Collective: every rank calls it for every global id, contributing zeros for patches it does not hold. | |
| pure logical function | m_ibm::f_needs_centroid_offset (patch) |
| A moving airfoil or STL (geometries 4, 5, 11, 12) moves its centroid to the centre of mass and keeps a centroid_offset. | |
| impure subroutine | m_ibm::s_restore_centroid_offsets (t_init) |
| On restart, replace the re-measured centroid offsets with the ones the run was using (restart_data/ib_offset_<step>.dat, written with each checkpoint), and re-place kinematics-driven bodies about them. No file: the measured offsets stand. | |
| subroutine | m_ibm::s_compute_moment_of_inertia (patch, axis, moment) |
| Computes the moment of inertia for an immersed boundary. | |
| subroutine | m_ibm::s_wrap_periodic_ibs () |
| Wrap immersed boundary positions across periodic domain boundaries. | |
| impure subroutine | m_ibm::s_setup_ib_force_comm () |
| Builds the ascending list of distinct neighborhood ranks used by s_communicate_ib_forces and allocates its buffers once. The ascending order is the fixed order in which an owner adds its neighbors' contributions. | |
| integer function | m_ibm::f_ib_nbr_slot (rank) |
| Position of rank in ib_nbrs. | |
| subroutine | m_ibm::s_communicate_ib_forces (forces, torques) |
| Reduces forces and torques so every rank holding an IB ends with bit-identical values. Each IB has one owner, the rank whose subdomain holds its centroid (patch_ibowner_rank, stamped at handoff). Phase 1: each rank sends its nonzero partials to the owner, which adds them to its own partial in ascending source-rank order once all have arrived. Phase 2: each owner sends its totals to every neighborhood rank, which overwrite their partials. One nonblocking message per neighbor per phase. | |
| subroutine | m_ibm::s_handoff_ib_ownership () |
| Swaps ownership of IBs and passes ownership of IBs to neighbor processors. | |
| impure subroutine | m_ibm::s_check_every_patch_marked () |
| Abort if any namelist immersed boundary marks no cell anywhere. The owning rank is chosen by the patch centroid, so an STL placed with model_translate while its centroid is left elsewhere is handed to ranks that do not hold it, and it vanishes silently. Cloud particles are skipped: a particle's centroid is its body, so it cannot vanish this way. | |
| impure subroutine, public | m_ibm::s_finalize_ibm_module () |
| Finalize the IBM module. | |
| subroutine | m_ibm::s_surface_species_residual (pres, t_s, d, ys_ip, ys_s, w_species, r_species, omega_s, mdot_s) |
| Species flux residual at a heterogeneous reacting surface. | |
| subroutine | m_ibm::s_surface_energy_residual (pres, t_ip, t_s, d, ys_s, r_energy) |
| Surface energy residual: gas-side conduction balances heterogeneous reaction heat. Radiation and solid-side conduction are omitted. | |
| subroutine | m_ibm::s_surface_residual (pres, t_ip, t_s, d, ys_ip, ys_s, w_species, k_bath, solve_temperature, flux_scale, energy_scale, r, r_species, omega_s, mdot_s) |
| Assemble the Newton residual for Ns species, with temperature appended only when it is solved. k_bath selects which species balance the sum constraint displaces; see s_solve_surface for why it must be the most abundant one. | |
| subroutine | m_ibm::s_blend_ghost_state (t_ip, t_s, ys_ip, ys_s, t_g, ys_g) |
| Ghost state for an immersed surface, from the image-point state and the surface state. | |
| subroutine | m_ibm::s_pick_bath_species (ys_s, k_bath) |
| Index of the most abundant species, i.e. the balance the sum constraint displaces. | |
| subroutine | m_ibm::s_solve_surface (pres, t_ip, t_wall, d, ys_ip, w_species, thermal_bc, ys_s, t_s, mdot_s, converged) |
| Newton solve for a reacting surface. thermal_bc=0: zero-normal-gradient T; 1: prescribed T; 2: energy balance. | |
| subroutine | m_ibm::s_solve_surface_linear_system (a, b, x, nsolve, success) |
| Small dense linear solve with partial pivoting for the local surface Newton system. | |
Variables | |
| integer | m_ibm::n_surface_not_converged = 0 |
| Ghost points at which the reacting-surface Newton solve did not reach its tolerance, so the point fell back to a chemically inert wall (keeping a prescribed Twall). Counted because that fallback is otherwise indistinguishable from a surface mechanism that simply does nothing, and a run can look converged while the surface chemistry never engaged. | |
| integer | m_ibm::n_surface_ill_posed = 0 |
| Ghost points whose levelset distance was not positive, leaving the surface gradient (X_IP - X_s)/d undefined. Separate from the counter above because this is a grid/geometry degeneracy, not a kinetics failure. | |
| type(integer_field), public | m_ibm::ib_markers |
| type(integer_field) | m_ibm::corrected_gps |
| 1 at ghost points whose image point reaches the fluid, 0 elsewhere. Ghost points with a buried image point average over these neighbors, and never over each other. | |
| type(ghost_point), dimension(:), allocatable | m_ibm::ghost_points |
| integer | m_ibm::num_gps |
| Number of ghost points. | |
| logical | m_ibm::moving_immersed_boundary_flag |
| logical | m_ibm::centroid_offsets_active = .false. |
| some patch is a moving airfoil or STL, which carries a centroid_offset | |
| integer, parameter, private | m_ibm::ib_rec_len = 7 |
| gbl_patch_id, force(1:3), torque(1:3) | |
| integer, private | m_ibm::n_ib_nbrs = 0 |
| integer, dimension(:), allocatable, private | m_ibm::ib_nbrs |
| integer, dimension(:), allocatable, private | m_ibm::ib_send_counts |
| integer, dimension(:), allocatable, private | m_ibm::ib_reqs |
| integer, dimension(:,:), allocatable, private | m_ibm::ib_stats |
| real(wp), dimension(:,:), allocatable, private | m_ibm::ib_send_bufs |
| real(wp), dimension(:,:), allocatable, private | m_ibm::ib_recv_bufs |
Contains module m_ibm.
Definition in file m_ibm.fpp.f90.