.. _auto_dataset_pick_curves_calibration:

*PICK CALIBRATION CURVES FROM DATASET*
======================================

Selects one representative stress-strain curve per (state, strain_rate) bucket from a dataset of physical-test curves using Pareto-front scoring with a non-crossing constraint between adjacent strain rates within each state. Use this worker to reduce a raw coupon-test dataset down to the calibration-ready curve set for material card authoring.

When to use
-----------

Classification: **process**.

Tagged: ``bucket_selection``, ``calibration``, ``coupon``, ``curve_selection``, ``material``, ``non_crossing``, ``outlier_removal``, ``pareto``.

Inputs
------

.. list-table::
   :header-rows: 1
   :widths: 20 20 20 20 20 20

   * - Label
     - ID
     - Type
     - Default
     - Required
     - Description
   * - Dataset Of Test Curves
     - dataset
     - dataset
     - —
     - ✓
     - Input dataset of physical test curves; each row must contain at minimum a state label, a strain-rate value, and a stress-strain curve cell.
   * - State Column
     - state_column
     - scalar
     - state
     - 
     - Name of the dataset column that identifies the deformation state (e.g. tension, compression, shear); defaults to 'state'.
   * - Strain Rate Column
     - strain_rate_column
     - scalar
     - strain_rate
     - 
     - Name of the numeric dataset column that holds the strain-rate value for each test record; defaults to 'strain_rate'.
   * - Curve Column
     - curve_column
     - scalar
     - eng_stress_strain
     - 
     - Name of the dataset column whose cells contain the engineering stress-strain curve objects to be evaluated; defaults to 'eng_stress_strain'.
   * - Pareto Selection Criteria
     - pareto_criteria
     - dataset
     - *(complex)*
     - 
     - Table defining which curve features (slope, xmax, ymax, xatymax, integral, …) participate in Pareto-front scoring and whether each should be minimised or maximised; leave at default (slope/xmax/ymax all minimised) for standard conservative curve selection.
   * - Type Of Data To Use
     - data_type
     - scalar
     - median
     - 
     - Controls whether Pareto feature scores are measured relative to the per-bucket median ('median', default) or to raw feature values ('raw'); use 'raw' when bucket populations are very small.
   * - Remove Outliers
     - remove_outliers
     - scalar
     - ymax
     - 
     - Pre-pass outlier removal strategy applied before Pareto scoring: 'none', 'xmax' (filter by failure strain), 'ymax' (filter by peak stress, default), or 'both'.
   * - Outlier Tolerance
     - outlier_tolerance
     - scalar
     - 0.5
     - 
     - Fractional deviation threshold used by the outlier-removal pre-pass; 0.5 means a curve is flagged if its feature deviates more than 50 % from the bucket median (default 0.5).
   * - Non-crossing Check: Minimum X
     - non_crossing_check_min_x
     - scalar
     - auto
     - 
     - Lowest x at which to start checking for crossings. 'auto' = max(curve_a.xmin, curve_b.xmin)
   * - When A (state, rate) Bucket Has No Surviving Candidate
     - on_empty_bucket
     - scalar
     - omit
     - 
     - Behaviour when a (state, strain_rate) bucket contains no valid curves after outlier removal: 'omit' (default) skips the bucket silently, other values may raise a warning or error.
   * - Cross-rate Envelope (Spread)
     - pick_envelope
     - scalar
     - upper
     - 
     - Cross-rate spread objective. Default 'upper' targets the upper envelope of the candidate cloud at every rate so picks demonstrate strain-rate sensitivity instead of clustering (verified against Ashutosh's manual selection on the ABS PBUV cloud, JIRA DW-2900). 'lower' flips to ymax=min everywhere. 'auto_by_rate' currently behaves like 'upper'; reserved for future per-rate auto-detection. 'none' is legacy v1 — apply user criteria uniformly. Only the ymax criterion direction is rewritten; slope and xmax rows pass through unchanged.
   * - Final Remove-Intersections Pass
     - final_remove_intersections
     - scalar
     - yes
     - 
     - After per-bucket selection and the non-crossing walk, lift residual crossings between adjacent strain-rate picks via CurvesTransformer::remove_intersections. Default 'yes' (matches the post-processing path used downstream of typical material-card calibration). Mirrors the toggle used by extrapolate / sync_last_slope / interpolate_curves_*.
   * - Final Removal Method
     - final_removal_method
     - scalar
     - offset
     - 
     - Removal-method passthrough to CurvesTransformer::remove_intersections; only consulted when 'Final Remove-Intersections Pass' is not 'no'. Default 'offset'.

Outputs
-------

.. list-table::
   :header-rows: 1
   :widths: 20 20 20 20

   * - Label
     - ID
     - Type
     - Description
   * - Picked Curves
     - picked_curves
     - dataset
     - Dataset containing exactly one selected stress-strain curve per (state, strain_rate) bucket, with all original metadata columns preserved alongside the chosen curve cell.
   * - Scatter Statistics
     - scatter_stats
     - dataset
     - Dataset of per-bucket descriptive statistics (e.g. feature means, medians, scatter metrics) computed during Pareto scoring, useful for auditing selection quality and documenting test scatter.

Disciplines
-----------

- data.curve.pair
- data.dataset.transform
- engineering.material.calibration
- engineering.material.characterization
- engineering.material.specimen

.. raw:: html

   <hr style="margin-top:2em">
   <p style="font-size:11px;color:#888">
   Auto-generated from <code>transformation</code> schema. Worker id: <code>dataset_pick_curves_calibration</code>. Schema hash: <code>c28e56473f97</code>. Hand-curated docs in <code>workerexamples/</code> override this page when present.
   </p>