Drop rebinning of meanstress transform histogram by johannes-mueller · Pull Request #233 · boschresearch/pylife

johannes-mueller · 2026-03-19T15:01:14Z

Fix #217

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

johannes-mueller · 2026-03-19T15:01:38Z

cc @Steve-WV

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

Copilot

Pull request overview

This PR addresses Issue #217 by changing mean stress transformation of 2D rainflow histograms to avoid rebinning (which can blur results), and updates tests/docs and supporting accessors accordingly.

Changes:

Reworks histogram mean stress transformation to preserve bin boundaries instead of rebinned, continuous 1D bins.
Adds helper metadata on load collectives/histograms to identify the stress-axis levels/columns.
Updates and expands test coverage (including additional-index scenarios) and refreshes related docs/examples.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 11 comments.

Show a summary per file

File	Description
`src/pylife/strength/meanstress.py`	Implements non-rebinned histogram transformation and expands documentation/examples.
`src/pylife/stress/collective/load_histogram.py`	Tracks histogram stress-axis index levels via a new `index_levels` property.
`src/pylife/stress/collective/load_collective.py`	Tracks load-collective stress-axis columns via a new `columns` property.
`tests/strength/test_meanstress_matrix.py`	Updates expectations for empty transformations to match new index structure.
`tests/strength/test_meanstress.py`	Removes/updates tests around rebinning and adds new scenarios validating non-rebinned behavior.
`demos/lifetime_calc.ipynb`	Minor demo adjustments (filtering/variable split) and notebook metadata update.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-20T13:48:38Z

+        have_additional_indeces = len(transformed.index.names) > 2
+        if have_additional_indeces:
+            transformed.index = transformed.index.droplevel(self.index_levels)


Typo in variable name: have_additional_indeces should be have_additional_indices for clarity and to avoid propagating the misspelling.

Copilot · 2026-03-20T13:48:38Z

+    print(expected)
+    print(transformed.to_pandas())
+


Please remove debug print() calls from tests; they add noise to test output and can slow down CI runs.

Copilot · 2026-03-20T13:48:39Z

+    print(expected)
+    print(transformed.to_pandas())
+


Please remove debug print() calls from tests; they add noise to test output and can slow down CI runs.

Copilot · 2026-03-20T13:48:39Z

+    print(expected)
+    print(transformed.to_pandas())
+


Please remove debug print() calls from tests; they add noise to test output and can slow down CI runs.

Copilot · 2026-03-20T13:48:39Z

+    print(transformed.to_pandas())
+    print(expected)
+
+    pd.testing.assert_series_equal(transformed.to_pandas(), expected)
+


Please remove debug print() calls from tests; they add noise to test output and can slow down CI runs.

Suggested change

print(transformed.to_pandas())

print(expected)

pd.testing.assert_series_equal(transformed.to_pandas(), expected)

pd.testing.assert_series_equal(transformed.to_pandas(), expected)

Copilot · 2026-03-20T13:48:40Z

+    print(transformed.to_pandas())
+    print(expected)
+


Please remove debug print() calls from tests; they add noise to test output and can slow down CI runs.

Suggested change

print(transformed.to_pandas())

print(expected)

Copilot · 2026-03-20T13:48:40Z

+    print(expected)
+    print(transformed.to_pandas())
+


Please remove debug print() calls from tests; they add noise to test output and can slow down CI runs.

Copilot · 2026-03-20T13:48:40Z

+    print(expected)
+    print(transformed.to_pandas())
+


Please remove debug print() calls from tests; they add noise to test output and can slow down CI runs.

Copilot · 2026-03-20T13:48:40Z

+        orig_left = pd.DataFrame({"range": range_left, "mean": mean_left})
+        orig_right = pd.DataFrame({"range": range_right, "mean": mean_right})
+
+        transformed_left = transformer.transform(orig_left, R_goal)
+        transformed_right = transformer.transform(orig_right, R_goal)
+
+        transformed = self._obj.to_frame()
+
+        have_additional_indeces = len(transformed.index.names) > 2
+        if have_additional_indeces:
+            transformed.index = transformed.index.droplevel(self.index_levels)
+
+        range = pd.DataFrame({0: transformed_left["range"], 1: transformed_right["range"]})
+        transformed["range"] = pd.IntervalIndex.from_arrays(
+            range.min(axis=1), range.max(axis=1), name="range"
+        )
+        mean = pd.DataFrame({0: transformed_left["mean"], 1: transformed_right["mean"]})
+        transformed["mean"] = pd.IntervalIndex.from_arrays(
+            mean.min(axis=1), mean.max(axis=1), name="mean"


In _perform_transformation, the new interval boundaries are derived only from use_class_left() (left/left corner) and use_class_right() (right/right corner). For from/to histograms (and even for range/mean bins when sensitivity depends on mean), extrema of transformed range/mean over the original 2D bin can occur at mixed corners (left/right or right/left). This can produce transformed bins that do not correspond to the full original bin limits and can under- or over-estimate the resulting intervals. Consider transforming all corner combinations of the original bin edges (4 corners) and taking min/max to build the resulting IntervalIndex so the resulting bins truly correspond to the original 2D bins.

Suggested change

orig_left = pd.DataFrame({"range": range_left, "mean": mean_left})

orig_right = pd.DataFrame({"range": range_right, "mean": mean_right})

transformed_left = transformer.transform(orig_left, R_goal)

transformed_right = transformer.transform(orig_right, R_goal)

transformed = self._obj.to_frame()

have_additional_indeces = len(transformed.index.names) > 2

if have_additional_indeces:

transformed.index = transformed.index.droplevel(self.index_levels)

range = pd.DataFrame({0: transformed_left["range"], 1: transformed_right["range"]})

transformed["range"] = pd.IntervalIndex.from_arrays(

range.min(axis=1), range.max(axis=1), name="range"

)

mean = pd.DataFrame({0: transformed_left["mean"], 1: transformed_right["mean"]})

transformed["mean"] = pd.IntervalIndex.from_arrays(

mean.min(axis=1), mean.max(axis=1), name="mean"

# Build all four corner combinations of the original 2D bins

orig_ll = pd.DataFrame({"range": range_left, "mean": mean_left})

orig_rr = pd.DataFrame({"range": range_right, "mean": mean_right})

orig_lr = pd.DataFrame({"range": range_left, "mean": mean_right})

orig_rl = pd.DataFrame({"range": range_right, "mean": mean_left})

# Concatenate corners with a MultiIndex so we can aggregate per original bin

orig_corners = pd.concat(

[orig_ll, orig_rr, orig_lr, orig_rl],

keys=["ll", "rr", "lr", "rl"],

names=["corner", "bin"],

)

transformed_corners = transformer.transform(orig_corners, R_goal)

transformed = self._obj.to_frame()

have_additional_indeces = len(transformed.index.names) > 2

if have_additional_indeces:

transformed.index = transformed.index.droplevel(self.index_levels)

# For each original bin, determine the extrema of the transformed values

range_wide = transformed_corners["range"].unstack(0)

transformed["range"] = pd.IntervalIndex.from_arrays(

range_wide.min(axis=1), range_wide.max(axis=1), name="range"

)

mean_wide = transformed_corners["mean"].unstack(0)

transformed["mean"] = pd.IntervalIndex.from_arrays(

mean_wide.min(axis=1), mean_wide.max(axis=1), name="mean"

Copilot · 2026-03-20T13:48:40Z

+        goodman: pd.Series or pd.DataFrame
+           The meanstress sensitivity data needs `M` and optionally `M2`
+
+        R_goal: float
+           The R-value to transform to
+
+        Returns
+        -------
+        transformed_collective: LoadHistogram
+            The transformed load histogram


The fkm_goodman() docstring parameter list says goodman, but the function argument is named haigh. Also the Returns section labels the return as transformed_collective even though the method returns a LoadHistogram. Please align the parameter/return names in the docstring with the actual signature/type, and document that the transformed histogram bins can be non-contiguous (no rebinning) and that users may optionally rebin afterwards for visualization (with an accuracy tradeoff).

Suggested change

goodman: pd.Series or pd.DataFrame

The meanstress sensitivity data needs `M` and optionally `M2`

R_goal: float

The R-value to transform to

Returns

-------

transformed_collective: LoadHistogram

The transformed load histogram

haigh: pd.Series or pd.DataFrame

The meanstress sensitivity data; needs column ``M`` and optionally ``M2``.

R_goal: float

The R-value to transform to.

Returns

-------

LoadHistogram

The transformed load histogram as a :class:`~pylife.stress.collective.LoadHistogram`.

Note that the resulting amplitude/mean bins can be non-contiguous because no

rebinning is performed during the transformation. You may optionally rebin

the histogram afterwards for visualization, accepting the corresponding

loss of accuracy.

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

Copilot

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 13 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-09T06:34:59Z

-            res["cycles"] = original_cycles["cycles"]
+        for col in collective:
+            if col not in coll.columns:
+                res[col] = collective[col]


In broadcasted HaighDiagram use-cases (e.g. haigh indexed by element_id), broadcasted_coll has a different (expanded) index than the original collective. Copying extra columns from collective[col] will misalign and fill with NaN. Copy from broadcasted_coll (or broadcast the column) so additional data like cycles is preserved under broadcasting.

Suggested change

res[col] = collective[col]

res[col] = broadcasted_coll[col]

Copilot · 2026-04-09T06:34:59Z

+        range_left = 2.0 * self.use_class_left().amplitude.reset_index(drop=True)
+        range_right = 2.0 * self.use_class_right().amplitude.reset_index(drop=True)
+        mean_left = self.use_class_left().meanstress.reset_index(drop=True)
+        mean_right = self.use_class_right().meanstress.reset_index(drop=True)
+
+        range_index = self.amplitude_histogram.index
+        range_left = 2.0 * range_index.left.to_series().reset_index(drop=True)
+        range_right = 2.0 * range_index.right.to_series().reset_index(drop=True)
+
+        orig_lele = pd.DataFrame({"range": range_left, "mean": mean_left})
+        orig_lere = pd.DataFrame({"range": range_left, "mean": mean_right})
+        orig_rele = pd.DataFrame({"range": range_right, "mean": mean_left})
+        orig_rere = pd.DataFrame({"range": range_right, "mean": mean_right})
+
+        transformed_lele = transformer.transform(orig_lele, R_goal)
+        transformed_lere = transformer.transform(orig_lere, R_goal)
+        transformed_rele = transformer.transform(orig_rele, R_goal)
+        transformed_rere = transformer.transform(orig_rere, R_goal)


For histograms indexed by ('from','to'), the four corner points of each 2D bin are (from.left/to.left), (from.left/to.right), (from.right/to.left), (from.right/to.right). This code mixes range_left/right with mean_left/right, which are derived from different corner combinations, so it does not transform the actual bin corners and can yield incorrect (non-conservative or overly conservative) resulting intervals. Build the four corner collectives from the original index edges and transform those, then take min/max of the transformed range/mean.

Copilot · 2026-04-09T06:35:00Z

+            The transformed load histogram
+


The docstring for histogram meanstress transformation doesn’t mention that the resulting (range, mean) bins may be non-continuous now, nor that users may optionally rebin afterwards for visualization at the cost of accuracy (per issue #217 requirements). Please document this behavioral change here.

Suggested change

The transformed load histogram

The transformed load histogram. After the meanstress transformation,

the resulting ``(range, mean)`` interval bins may no longer be

continuous.

Notes

-----

If continuous bins are required afterwards, e.g. for visualization, the

transformed histogram can optionally be rebinned. Be aware that such a

rebinning is a post-processing step for presentation purposes and may

reduce the accuracy of the transformed histogram.

Copilot · 2026-04-09T06:35:00Z

+    def _perform_transformation(self, transformer, R_goal):
+        range_left = 2.0 * self.use_class_left().amplitude.reset_index(drop=True)
+        range_right = 2.0 * self.use_class_right().amplitude.reset_index(drop=True)
+        mean_left = self.use_class_left().meanstress.reset_index(drop=True)
+        mean_right = self.use_class_right().meanstress.reset_index(drop=True)
+
+        range_index = self.amplitude_histogram.index
+        range_left = 2.0 * range_index.left.to_series().reset_index(drop=True)
+        range_right = 2.0 * range_index.right.to_series().reset_index(drop=True)
+


range_left/range_right are computed from use_class_left/right() and then immediately overwritten from self.amplitude_histogram.index. This dead code makes the transformation logic harder to follow; please remove the unused assignments (and consider renaming range/have_additional_indeces to avoid shadowing the built-in and typos).

Copilot · 2026-04-09T06:35:00Z

+@pytest.mark.skip("We are no longer rebinning")
 @pytest.mark.parametrize("R_goal, expected", [  # all calculated by pencil on paper
    (-1., 2.0),
    (0., 4./3.),
    (-1./3., 8./5.),
    (1./3., 14./12.)
 ])
-def test_fkm_goodman_hist_range_mean_nonzero(R_goal, expected):
+def test_fkm_goodman_hist_range_mean_nonzero_binsize(R_goal, expected):
    rg = pd.IntervalIndex.from_breaks(np.linspace(0, 2, 25), closed='left')


Please avoid leaving skipped tests in the suite. If this test no longer matches the new non-rebinned behavior, it should be updated to assert the new binning semantics or removed entirely to keep coverage meaningful.

Copilot · 2026-04-09T06:35:01Z

+    )
+
+    print(transformed.to_pandas())
+    print(expected)
+
+    pd.testing.assert_series_equal(transformed.to_pandas(), expected)
+


Debug print() statements in tests add noise to CI output and make failures harder to read. Please remove these prints.

Copilot · 2026-04-09T06:35:01Z

+        name="cycles"
+    )
+    print(expected)
+    print(transformed.to_pandas())
+
+    pd.testing.assert_series_equal(transformed.to_pandas(), expected)
+


Debug print() statements in tests add noise to CI output and make failures harder to read. Please remove these prints.

Copilot · 2026-04-09T06:35:01Z

+        name="cycles"
+    )
+    print(expected)
+    print(transformed.to_pandas())
+
+    pd.testing.assert_series_equal(transformed.to_pandas(), expected)
+


Debug print() statements in tests add noise to CI output and make failures harder to read. Please remove these prints.

Copilot · 2026-04-09T06:35:01Z

+    )
+
+    print(transformed.to_pandas())
+    print(expected)
+
+    pd.testing.assert_series_equal(transformed.to_pandas(), expected)


Debug print() statements in tests add noise to CI output and make failures harder to read. Please remove these prints.

Copilot · 2026-04-09T06:35:02Z

+    print(transformed.to_pandas())
+    print(expected)
+


Debug print() statements in tests add noise to CI output and make failures harder to read. Please remove these prints.

Suggested change

print(transformed.to_pandas())

print(expected)

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

johannes-mueller added 3 commits March 19, 2026 16:00

No longer rebin meanstress transformed histograms

8da149e

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

Restrict lifetime_calc damage calculation to non-zero bins

c7524d5

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

Add docstrings to meanstress transform module

01f0982

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

johannes-mueller temporarily deployed to release March 19, 2026 15:01 — with GitHub Actions Inactive

johannes-mueller temporarily deployed to release March 19, 2026 15:07 — with GitHub Actions Inactive

Small cleanups

4972257

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

johannes-mueller temporarily deployed to release March 20, 2026 09:25 — with GitHub Actions Inactive

johannes-mueller temporarily deployed to release March 20, 2026 09:30 — with GitHub Actions Inactive

johannes-mueller requested a review from Copilot March 20, 2026 13:41

Copilot started reviewing on behalf of johannes-mueller March 20, 2026 13:41 View session

Copilot AI reviewed Mar 20, 2026

View reviewed changes

johannes-mueller added 3 commits March 24, 2026 15:53

Properly transform 2d load histogram squares

7a4561f

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

Fix amplitude calculation for from-to load histograms

637dc03

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

Improve variable names

570345c

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

johannes-mueller temporarily deployed to release March 24, 2026 15:06 — with GitHub Actions Inactive

johannes-mueller temporarily deployed to release March 24, 2026 15:11 — with GitHub Actions Inactive

johannes-mueller requested a review from Copilot April 9, 2026 06:28

Copilot started reviewing on behalf of johannes-mueller April 9, 2026 06:28 View session

Copilot AI reviewed Apr 9, 2026

View reviewed changes

Cleanups and docstring improvements

6ea19ff

Signed-off-by: Johannes Mueller <johannes.mueller4@de.bosch.com>

johannes-mueller temporarily deployed to release April 9, 2026 07:15 — with GitHub Actions Inactive

Merge branch 'develop' into 217-meanstress-transform-histogram

e2a4a40

johannes-mueller temporarily deployed to release April 9, 2026 07:17 — with GitHub Actions Inactive

johannes-mueller temporarily deployed to release April 9, 2026 07:22 — with GitHub Actions Inactive

johannes-mueller merged commit a5b5b2a into develop Apr 9, 2026
24 checks passed

johannes-mueller deleted the 217-meanstress-transform-histogram branch April 9, 2026 07:37

-            The transformed load histogram
+            The transformed load histogram. After the meanstress transformation,
+            the resulting ``(range, mean)`` interval bins may no longer be
+            continuous.
+        Notes
+        -----
+        If continuous bins are required afterwards, e.g. for visualization, the
+        transformed histogram can optionally be rebinned. Be aware that such a
+        rebinning is a post-processing step for presentation purposes and may
+        reduce the accuracy of the transformed histogram.

Conversation

johannes-mueller commented Mar 19, 2026

Uh oh!

johannes-mueller commented Mar 19, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone