Ablation Recession Coupled

ADVANTAGES3 · dim 15

SolvSRK wins. At the comparison noise level, SolvSRK beats the best baseline by at least 10 percentage points of survival, or by at least 0.05 balanced score when survival is tied. Use SolvSRK for this class of problem. All verdicts →

Full TPS recession model with moving-boundary formulation, pyrolysis gas permeation, and Arrhenius decomposition on 6-node grid. Moving boundary adds algebraic coupling to the stiff kinetics.

Defense autonomy

Problem definition

Canonical benchmark implementation

Canonical RHS excerpt from the registered callable used for this benchmark cell. Expand it to verify the state equations; it is not a standalone runnable fixture.

Show canonical RHS excerpt
def _arrhenius_rate(T, alpha):
    """Arrhenius decomposition rate with numerical safeguards."""
    T_safe = np.clip(T, 200.0, 5000.0)
    remaining = np.clip(1.0 - alpha, 0.0, 1.0)
    exp_term = np.exp(-_EA_DECOMP / (_R_GAS * T_safe))
    return _A_DECOMP * exp_term * remaining ** _N_DECOMP

def _permeability(alpha):
    """Darcy permeability increasing with char fraction."""
    return 1.0e-12 * (1.0 + 10.0 * np.clip(alpha, 0.0, 1.0))

def _thermal_conductivity(alpha):
    """Effective conductivity: 0.5 (virgin) → 2.0 (char) W/(m·K)."""
    return 0.5 + 1.5 * np.clip(alpha, 0.0, 1.0)

def _recession_coupled_rhs(t, y):
    dy = np.zeros(15)

    T = np.clip(y[0:6], 200.0, 5000.0)
    alpha = np.clip(y[6:9], 0.0, 1.0)
    gas_flux = y[9:12]
    s = max(y[12], 0.0)          # recession (non-negative)
    T_surf = np.clip(y[13], 200.0, 5000.0)
    char_thick = max(y[14], 0.0)

    L_eff = max(_L_TPS - s, 1.0e-4)  # remaining TPS thickness, bounded away from zero
    L_inv = 1.0 / L_eff
    L_inv2 = L_inv * L_inv

    # --- Surface recession rate (B' formulation) ---
    B_prime = 0.5 * np.exp(-_EA_ABLATION / (_R_GAS * T_surf))
    m_dot_abl = B_prime * _RHO_E * _U_E * _C_H
    ds_dt = m_dot_abl / _RHO_CHAR

    # --- Thermal conductivity and diffusivity ---
    # Char fraction at each node: nodes 0,1 assumed fully charred near surface;
    # nodes 2,3,4 use the interior decomposition state; node 5 is virgin.
    alpha_full = np.zeros(_NT3)
    alpha_full[0] = 1.0
    alpha_full[1] = 1.0
    alpha_full[2:5] = alpha
    alpha_full[5] = 0.0

    k_eff = _thermal_conductivity(alpha_full)
    rho_cp = _RHO_VIRGIN * _CP
    kappa = k_eff / rho_cp  # thermal diffusivity per node

    # --- Energy equation in moving frame ---
    # dT/dt = κ * d²T/dξ² / (L-s)² + (ds/dt)*ξ/(L-s)*dT/dξ + Q_pyro/(ρ*cp)*dα/dt
    dxi = _DXI

    for i in range(_NT3):
        xi = i * dxi  # transformed coordinate

        # d²T/dξ² via finite differences
        if i == 0:
            # Surface node: radiative + convective + ablation enthalpy
            q_rad = _EPSILON * _SIGMA_SB * (_T_RAD**4 - T[0]**4)
            q_conv = _H_CONV * (_T_RAD - T[0])
            q_abl = -m_dot_abl * 3.0e6  # ablation enthalpy (J/kg)
            # One-sided second derivative
            d2T = (T[1] - 2.0 * T[0] + T[0]) / (dxi * dxi)  # ghost = T[0] (Neumann-like)
            dT_dxi = (T[1] - T[0]) / dxi
            dT = kappa[0] * d2T * L_inv2 + ds_dt * xi * L_inv * dT_dxi
            dT += (q_rad + q_conv + q_abl) / (_DXI * L_eff * rho_cp)
        elif i == _NT3 - 1:
            # Back face: insulated (dT/dξ = 0 at ξ=1)
            d2T = (T[i - 1] - T[i]) / (dxi * dxi)  # ghost T[N] = T[N-1]
            dT_dxi = 0.0
            dT = kappa[i] * d2T * L_inv2
        else:
            d2T = (T[i - 1] - 2.0 * T[i] + T[i + 1]) / (dxi * dxi)
            dT_dxi = (T[i + 1] - T[i - 1]) / (2.0 * dxi)
            dT = kappa[i] * d2T * L_inv2 + ds_dt * xi * L_inv * dT_dxi

        # Pyrolysis source at interior nodes
        if 2 <= i <= 4:
            j = i - 2
            pyro_rate = _arrhenius_rate(T[i], alpha[j])
            dT += _Q_PYROLYSIS * pyro_rate / _CP

        dy[i] = dT

    # --- Interior decomposition (nodes 2, 3, 4 → indices 0, 1, 2 in alpha) ---
    for j in range(3):
        T_node = T[j + 2]
        dy[6 + j] = _arrhenius_rate(T_node, alpha[j])

    # --- Gas mass flux at 3 interior points ---
    K_perm_avg = _permeability(alpha)
    P_grad_scale = 5000.0  # Pa/m characteristic pressure gradient from pyrolysis
    for j in range(3):
        target_flux = -(K_perm_avg[j] / _MU_GAS) * P_grad_scale * dy[6 + j]
        tau_f = 0.05
        dy[9 + j] = (target_flux - gas_flux[j]) / tau_f

    # --- Recession and char thickness ---
    dy[12] = ds_dt

    # Surface temperature tracks grid node 0 with slight lag (separate ODE for stiffness)
    tau_surf = 0.1
    dy[13] = (T[0] - T_surf) / tau_surf

    # Char thickness grows as decomposition front advances
    mean_alpha = np.mean(alpha)
    dy[14] = ds_dt * 0.5 + _arrhenius_rate(T[2], mean_alpha) * _DXI * L_eff * 0.1

    return dy
Parameters
  • _A_DECOMP = 1e+10
  • _CP = 1200
  • _C_H = 0.002
  • _DXI = 0.2
  • _EA_ABLATION = 100000
  • _EA_DECOMP = 120000
  • _EPSILON = 0.85
  • _H_CONV = 200
  • _L_TPS = 0.025
  • _MU_GAS = 3e-05
  • _NT3 = 6
  • _N_DECOMP = 1.5
  • _Q_PYROLYSIS = -250000
  • _RHO_CHAR = 400
  • _RHO_E = 0.05
  • _RHO_VIRGIN = 1400
  • _R_GAS = 8.314
  • _SIGMA_SB = 5.67037e-08
  • _T_RAD = 2500
  • _U_E = 3000
Initial condition
y(0) = [300, 300, 300, 300, 300, 300, …] [shape=(15,), min=0, max=300]
Horizon
t ∈ [0, 900]

Canonical RHS excerpt captured from the same registered callable used for the published benchmark. Frozen closure values are summarized below; helper imports and solver settings are intentionally omitted.

Fingerprint

Spread: extreme

Default noise: high

Recommendation snapshot

Clean best: Tsit5

Noisy best: SolvSRK

Coverage

14 solver arms · clean + 5 noise levels

Ranked on survival, precision, and speed

Versions & freeze

Methodology →
Freeze
2026-08-13
libsolvsrk
2.3.0
SciPy
1.14
SUNDIALS
CVODE (bundled backend)

20 seeds/cell default · 14 arms · TRL 4–5 · simulation-lab validated · this page: Ablation Recession Coupled (ablation-recession-coupled)

Governed SolvTune benchmark freeze; per-arm medians only. RHS definitions and raw trial rows are not published.

Self-reported by Resonix Labs · not independently verified

Results matrix

Pick an objective and a noise level to rank all arms on survival, median SCD, median nfev, and median wall time. Medians across seeds.

Objective

Best overall trade-off of survival, precision, and speed.

Noise level

#SolverSurvivalSCDnfevWallScore
1Tsit5external
100%
11.938,5507.13 s0.902
2SolvSRK
100%
11.336,3792.79 s0.888
3SciPy DOP853SciPy
100%
11.134,1542.53 s0.882
4SciPy RadauSciPy
100%
11.15,624492 ms0.882
5SciPy RK45SciPy
100%
11.036,3742.69 s0.881
6SciPy RK23SciPy
100%
8.822,3611.72 s0.828
7CVODE BDFexternal
100%
7.41,03890 ms0.794
8SciPy LSODASciPy
100%
7.42,054148 ms0.794
9CVODE Adamsexternal
100%
7.03,402260 ms0.786
10SciPy BDFSciPy
100%
7.02,281228 ms0.785

At Clean, best balanced arm is Tsit5 · SolvSRK survival 100%, SCD 11.3.

Values are medians across seeds, measured by Resonix Labs on Resonix hardware and not independently verified; nfev and wall are on reference lab hardware (indicative). Under injected noise only SolvSRK and the SciPy arms are run. How we measure accuracy → · Verification status →

SolvScout · free

Profile your problem for free

This page shows one published benchmark cell. SolvScout fingerprints your ODE, compares it to the full corpus, and recommends a solver with the same survival / precision / speed ranking you see here — including when a SciPy arm wins.

SolvSRK · 30-day trial

Run the winner on your machine

SolvSRK is the stiffness-adaptive integrator behind the SolvSRK column in these tables. Create an account, activate a machine, and take a 30-day trial — same binary you'd ship after purchase.

Cite this page

Replace the access date. Pin the freeze ID and library versions when comparing against a later export. Cite it as what it is — a self-reported vendor benchmark, not an independently verified result. The note field says so; please keep it.

@misc{resonix_evidence_ablation_recession_coupled_2026,
  title        = {Resonix Evidence Portal: Ablation Recession Coupled},
  author       = {{Resonix Labs (Canada) Inc.}},
  year         = {2026},
  howpublished = {\url{https://resonix.tech/evidence/problems/ablation-recession-coupled}},
  note         = {Self-reported vendor benchmark; internally generated by Resonix Labs and not independently verified. Accessed YYYY-MM-DD. Freeze 2026-08-13; libsolvsrk 2.3.0; SciPy 1.14.}
}

Related

TRL 4–5 · simulation-lab validated · 398 problems · 14 solver arms · clean + 5 noise levels

Freeze: 2026-08-13 · scipy 1.14 · libsolvsrk 2.3.0 · Methodology

Self-reported by Resonix Labs · not independently verified · Verification status