Modes and Hooks

A mode file must contain exactly one of Interface, Snapshots, or Calibrate.

Interface

{
  "Interface": {
    "settings": { "timestep": 1e-4 },
    "output": { "frequency": 10, "temperature": false },
    "hooks": {
      "solidification": {
        "params": { "dT_err": 1e-3, "max_iter": 10 },
        "outputs": {
          "tSol": true, "G": true, "G_vector": true,
          "G_unit_vector": false, "V": true,
          "dTdt": true, "numMelt": true
        }
      },
      "meltpool": {
        "params": { "MP_Stats_Output": "SpaceAndTime" },
        "outputs": { "MP_Stats": true, "Col_Depth": true }
      },
      "cet": {
        "params": { "N0": 3e13, "n": 3, "a": 1.25e6 },
        "outputs": { "eqFrac": true }
      }
    }
  }
}

settings.timestep is required and is in seconds. output is required; frequency defaults to 0 and temperature to false. Frequency 0 suppresses periodic files but the final file path is still reached. The loop continues past the last scan segment until the tracked liquid region disappears.

Hook objects are optional. An empty or missing hook is disabled.

• solidification: controls liquidus crossing refinement and fields described in Solidification Quantities. These fields are cleared when a tracked point becomes liquid and are replaced when it solidifies again. numMelt counts solid-to-liquid transitions.
• meltpool: MP_Stats requests length, width, and (for a 3D domain) depth. MP_Stats_Output may be Space, Time, or SpaceAndTime. Col_Depth records interpolated depth.
• cet: requires N0, n, and a; eqFrac is a required boolean when the hook is configured.

Coupled RDF/SRDF calls force Interface file output and JSON hooks off, then enable the calculations required by the selected Stork container.

Snapshots

{
  "Snapshots": {
    "settings": {
      "tracking": "Interface",
      "times": ["10%", "50%", "100%"]
    },
    "output": { "temperature": true },
    "hooks": {
      "meltpool": {
        "outputs": { "MP_Stats": true, "Col_Depth": true }
      }
    }
  }
}

settings.times must contain at least one finite, nonnegative time and must be strictly increasing after percentage strings are expanded relative to the longest scan duration. tracking is None (default) or Interface.

• None evaluates every grid point and forces full-volume 3D output.
• Interface traces the melted region, reducing work and output volume.

The optional snapshot melt-pool hook records one statistics row per snapshot and/or maps column depth to snapshot output.

Calibrate

Calibrate searches multiplicative parameters $\alpha_c$ and $\beta_c$ around the input material. A candidate changes diffusivity and volumetric heat capacity as

\[\alpha'=\alpha_0\alpha_c,\qquad (\rho c_p)'=\frac{(\rho c_p)_0}{\beta_c}.\]

The output material keeps input density, uses $c_p’=(\rho c_p)’/\rho$, and sets $k’=\alpha’(\rho c_p)’$.

Line evaluator with grid search

{
  "Calibrate": {
    "evaluator": {
      "type": { "line": {
        "velocity": 0.8, "length": 0.0006,
        "target_length": 240e-6,
        "target_width": 160e-6, "target_depth": 55e-6
      }},
      "weights": { "length": 1, "width": 1, "depth": 1, "distance": 0 },
      "loss": "percent"
    },
    "optimizer": {
      "type": { "grid": {
        "points": 5, "iterations": 6,
        "initial_span": 1, "tolerance": 1e-6
      }},
      "output": {
        "file": "Material-Optimized.json",
        "history": "Calibration-History.csv",
        "print": "on"
      }
    }
  }
}

Path evaluator with gradient descent

{
  "Calibrate": {
    "evaluator": {
      "type": { "path": {
        "time": [0.0005, 0.0010, 0.0015],
        "weights": [1, 1, 2],
        "target_length": [180e-6, 230e-6, 260e-6],
        "target_width": [110e-6, 145e-6, 165e-6],
        "target_depth": [35e-6, 50e-6, 60e-6]
      }},
      "weights": { "length": 1, "width": 1, "depth": 1, "distance": 0.05 },
      "loss": "percent"
    },
    "optimizer": {
      "type": { "gradient_descent": {
        "finite_difference": 0.01,
        "learning_rate": 0.25,
        "max_step": 0.25,
        "tolerance": 1e-3,
        "max_iterations": 100
      }},
      "output": {
        "file": "Material-Path-Gradient.json",
        "history": "Calibration-Path-Gradient.csv",
        "print": "on"
      }
    }
  }
}

Evaluator type is exactly one of:

• line: positive velocity and length, with one or more of target_length, target_width, target_depth.
• path: arrays time, optional per-time weights, and target arrays such as target_width. It evaluates the configured input path.

Loss is percent or absolute. Optimizer type is grid (odd points >= 3) or gradient_descent (finite_difference, learning_rate, max_step, tolerance, and max_iterations). Configuration keys and string values are case-sensitive.