Skip to content

Commit

Permalink
Merge pull request #115 from boschresearch/107-load-histogram
Browse files Browse the repository at this point in the history 
Add `LoadCollective.histogram()`
  • Loading branch information
@johannes-mueller
johannes-mueller authored Oct 10, 2024
2 parents e971770 + de1d2ee commit 5b9e004
Show file tree
Hide file tree
Showing 3 changed files with 377 additions and 25 deletions.
6 changes: 6 additions & 0 deletions CHANGELOG.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -5,9 +5,15 @@ In this file noteworthy changes of new releases of pyLife are documented since

## unreleased

### New features

* New method `LoadCollective.histogram()` (#107)

### Improvements

* Sanitize checks for Wöhler analysis (#108)
* Error messages when odbclient gets unsupported element types (#64)
* Improved documentation

### Bugfixes

Expand Down
201 changes: 191 additions & 10 deletions src/pylife/stress/collective/load_collective.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -17,14 +17,17 @@
__author__ = "Johannes Mueller"
__maintainer__ = __author__

from .abstract_load_collective import AbstractLoadCollective
from .load_histogram import LoadHistogram
import warnings

import pandas as pd
import numpy as np

from pylife import PylifeSignal

from .abstract_load_collective import AbstractLoadCollective
from .load_histogram import LoadHistogram


@pd.api.extensions.register_dataframe_accessor('load_collective')
class LoadCollective(PylifeSignal, AbstractLoadCollective):
"""A Load collective.
Expand Down Expand Up @@ -169,34 +172,212 @@ def shift(self, diffs):
return obj.load_collective

def range_histogram(self, bins, axis=None):
"""Calculate the histogram of range values along a given axis.
"""Calculate the histogram of cycles for range intervals along a given axis.
Parameters
----------
bins : int, sequence of scalars or pd.IntervalIndex
The bins of the histogram to be calculated
axis : str, optional
The index axis along which the histogram is calculated. If missing
the histogram is calculated over the whole collective.
Returns
-------
range histogram : :class:`~pylife.pylife.stress.LoadHistogram`
axis : str, optional
The index axis along which the histogram is calculated. If missing
the histogram is calculated over the whole collective.
Note
----
This resulting histogram does not contain any information on the mean
stress. Neither does it perform any kind of mean stress transformation
See also
--------
histogram
Examples
--------
Calculate a range histogram of a simple load collective
>>> df = pd.DataFrame(
... {'range': [1.0, 2.0, 1.0, 2.0, 1.0], 'mean': [0, 0, 0, 0, 0]},
... columns=['range', 'mean'],
... )
>>> df.load_collective.range_histogram([0, 1, 2, 3]).to_pandas()
range
(0, 1] 0
(1, 2] 3
(2, 3] 2
Name: cycles, dtype: int64
Calculate a range histogram of a load collective collection for
multiple nodes. The axis along which to aggregate the histogram is
given as ``cycle_number``.
>>> element_idx = pd.Index([10, 20, 30], name='element_id')
>>> cycle_idx = pd.Index([0, 1, 2], name='cycle_number')
>>> index = pd.MultiIndex.from_product((element_idx, cycle_idx))
>>> df = pd.DataFrame({
... 'range': [1., 2., 2., 0., 1., 2., 1., 1., 2.],
... 'mean': [0, 0, 0, 0, 0, 0, 0, 0, 0]
... }, columns=['range', 'mean'], index=index)
>>> h = df.load_collective.range_histogram([0, 1, 2, 3], 'cycle_number')
>>> h.to_pandas()
element_id range
10 (0, 1] 0
(1, 2] 1
(2, 3] 2
20 (0, 1] 1
(1, 2] 1
(2, 3] 1
30 (0, 1] 0
(1, 2] 2
(2, 3] 1
Name: cycles, dtype: int64
"""
def make_histogram(group):
cycles, intervals = np.histogram(group * 2., bins)
idx = pd.IntervalIndex.from_breaks(intervals, name='range')
return pd.Series(cycles, index=idx, name='cycles')

if isinstance(bins, pd.IntervalIndex):
if isinstance(bins, pd.IntervalIndex) or isinstance(bins, pd.arrays.IntervalArray):
bins = np.append(bins.left[0], bins.right)

if axis is None:
return LoadHistogram(make_histogram(self.amplitude))

result = pd.Series(self.amplitude
.groupby(self._obj.index.droplevel(axis).names)
.apply(make_histogram), name='cycles')
result = pd.Series(
self.amplitude.groupby(self._levels_from_axis(axis)).apply(
make_histogram
),
name='cycles',
)

return LoadHistogram(result)

def histogram(self, bins, axis=None):
"""Calculate the histogram of cycles along a given axis.
Parameters
----------
bins : int, sequence of scalars or pd.IntervalIndex
The bins of the histogram to be calculated
axis : str, optional
The index axis along which the histogram is calculated. If missing
the histogram is calculated over the whole collective.
Returns
-------
range histogram : :class:`~pylife.pylife.stress.LoadHistogram`
See also
--------
range_histogram
Examples
--------
Calculate a range histogram of a simple load collective
>>> df = pd.DataFrame(
... {'range': [1.0, 2.0, 1.0, 2.0, 1.0], 'mean': [0.5, 1.5, 1.0, 1.5, 0.5]},
... columns=['range', 'mean'],
... )
>>> df.load_collective.histogram([0, 1, 2, 3]).to_pandas()
range mean
(0, 1] (0, 1] 0.0
(1, 2] 0.0
(2, 3] 0.0
(1, 2] (0, 1] 2.0
(1, 2] 1.0
(2, 3] 0.0
(2, 3] (0, 1] 0.0
(1, 2] 2.0
(2, 3] 0.0
Name: cycles, dtype: float64
Calculate a range histogram of a load collective collection for
multiple nodes. The axis along which to aggregate the histogram is
given as ``cycle_number``.
>>> element_idx = pd.Index([10, 20], name='element_id')
>>> cycle_idx = pd.Index([0, 1, 2], name='cycle_number')
>>> index = pd.MultiIndex.from_product((element_idx, cycle_idx))
>>> df = pd.DataFrame({
... 'range': [1., 2., 2., 0., 1., 2.],
... 'mean': [0.5, 1.0, 1.0, 0.0, 1.0, 1.5]
... }, columns=['range', 'mean'], index=index)
>>> h = df.load_collective.histogram([0, 1, 2, 3], 'cycle_number')
>>> h.to_pandas()
element_id range mean
10 (0, 1] (0, 1] 0.0
(1, 2] 0.0
(2, 3] 0.0
(1, 2] (0, 1] 1.0
(1, 2] 0.0
(2, 3] 0.0
(2, 3] (0, 1] 0.0
(1, 2] 2.0
(2, 3] 0.0
20 (0, 1] (0, 1] 1.0
(1, 2] 0.0
(2, 3] 0.0
(1, 2] (0, 1] 0.0
(1, 2] 1.0
(2, 3] 0.0
(2, 3] (0, 1] 0.0
(1, 2] 1.0
(2, 3] 0.0
Name: cycles, dtype: float64
"""
def make_histogram(group):
cycles, range_bins, mean_bins = np.histogram2d(
group["range"], group["meanstress"], bins
)

return pd.Series(
cycles.ravel(),
name="cycles",
index=pd.MultiIndex.from_product(
[
pd.IntervalIndex.from_breaks(range_bins),
pd.IntervalIndex.from_breaks(mean_bins),
],
names=["range", "mean"],
),
)

range_mean = pd.DataFrame(
{'range': self.amplitude * 2, 'meanstress': self.meanstress},
index=self._obj.index,
)

if isinstance(bins, pd.IntervalIndex) or isinstance(bins, pd.arrays.IntervalArray):
bins = np.append(bins.left[0], bins.right)

if axis is None:
return LoadHistogram(make_histogram(range_mean))

# TODO: Warning filter can be dropped as soon as python-3.8 support is dropped
with warnings.catch_warnings():
warnings.simplefilter("ignore", category=FutureWarning)
result = pd.Series(
range_mean.groupby(self._levels_from_axis(axis))
.apply(make_histogram)
.stack(['range', 'mean']),
name="cycles",
)

return LoadHistogram(result)

def _levels_from_axis(self, axis):
return [lv for lv in self._obj.index.names if lv not in [axis] and lv is not None]
Loading

0 comments on commit 5b9e004

Please sign in to comment.