Use automatic binning of FKMNonLinearDetector

Signed-off-by: Johannes Mueller <[email protected]>
boschresearch · Dec 13, 2024 · 603a28f · 603a28f
1 parent 190e491
commit 603a28f
Show file tree

Hide file tree

Showing 3 changed files with 17 additions and 55 deletions.
diff --git a/src/pylife/materiallaws/notch_approximation_law_seegerbeste.py b/src/pylife/materiallaws/notch_approximation_law_seegerbeste.py
@@ -77,7 +77,7 @@ def stress(self, load, *, rtol=1e-4, tol=1e-4):
             The resulting stress
         '''
         # initial value as given by correction document to FKM nonlinear
-        x0 = load * (1 - (1 - 1/self._K_p)/1000)
+        x0 = np.asarray(load * (1 - (1 - 1/self._K_p)/1000))
 
         # suppress the divergence warnings
         with warnings.catch_warnings():
@@ -96,7 +96,8 @@ def stress(self, load, *, rtol=1e-4, tol=1e-4):
             # or (value, converged, zero_der) for vector-valued invocation
 
         # only for multiple points at once, if some points diverged
-        if len(stress) == 3 and not stress[1].all():
+        multidim = len(x0.shape) > 1 and x0.shape[1] > 1
+        if multidim and not stress[1].all():
             stress = self._stress_fix_not_converged_values(stress, load, x0, rtol, tol)
 
         return stress[0]
@@ -461,7 +462,7 @@ def _stress_fix_not_converged_values(self, stress, load, x0, rtol, tol):
         '''For the values that did not converge in the previous vectorized call to optimize.newton,
         call optimize.newton again on the scalar value. This usually finds the correct solution.'''
 
-        indices_diverged = [index for index, is_converged in enumerate(stress[1]) if not is_converged]
+        indices_diverged = np.where(~stress[1].all(axis=1))[0]
         x0_array = np.asarray(x0)
         load_array = np.asarray(load)
 
@@ -471,13 +472,13 @@ def _stress_fix_not_converged_values(self, stress, load, x0, rtol, tol):
             load_diverged = load_array[index_diverged]
             result = optimize.newton(
                 func=self._stress_implicit,
-                x0=x0_diverged,
+                x0=np.asarray(x0_diverged),
                 args=([load_diverged]),
                 full_output=True,
                 rtol=rtol, tol=tol, maxiter=50
             )
 
-            if result[1].converged:
+            if result.converged.all():
                 stress[0][index_diverged] = result[0]
         return stress
 

diff --git a/src/pylife/strength/fkm_nonlinear/assessment_nonlinear_standard.py b/src/pylife/strength/fkm_nonlinear/assessment_nonlinear_standard.py
@@ -389,27 +389,21 @@ def _compute_component_woehler_curves(assessment_parameters):
     return assessment_parameters, component_woehler_curve_P_RAM, component_woehler_curve_P_RAJ
 
 
-def _compute_hcm_RAM(assessment_parameters, scaled_load_sequence, maximum_absolute_load):
+def _compute_hcm_RAM(assessment_parameters, scaled_load_sequence, maximum_absolute_):
     """Perform the HCM rainflow counting with the extended Neuber notch approximation.
     The HCM algorithm is executed twice, as described in the FKM nonlinear guideline."""
 
     # initialize notch approximation law
     E, K_prime, n_prime, K_p = assessment_parameters[["E", "K_prime", "n_prime", "K_p"]]
     extended_neuber = pylife.materiallaws.notch_approximation_law.ExtendedNeuber(E, K_prime, n_prime, K_p)
 
-    # wrap the notch approximation law by a binning class, which precomputes the values
-    extended_neuber_binned = (
-        pylife.materiallaws.notch_approximation_law.NotchApproxBinner(
-            extended_neuber
-        ).initialize(maximum_absolute_load)
-    )
-
     # create recorder object
     recorder = pylife.stress.rainflow.recorders.FKMNonlinearRecorder()
 
     # create detector object
     detector = pylife.stress.rainflow.fkm_nonlinear.FKMNonlinearDetector(
-        recorder=recorder, notch_approximation_law=extended_neuber_binned, binner=None)
+        recorder=recorder, notch_approximation_law=extended_neuber
+    )
 
     # perform HCM algorithm, first run
     detector.process_hcm_first(scaled_load_sequence)
@@ -418,7 +412,7 @@ def _compute_hcm_RAM(assessment_parameters, scaled_load_sequence, maximum_absolu
     # perform HCM algorithm, second run
     detector.process_hcm_second(scaled_load_sequence)
 
-    return detector_1st, detector, extended_neuber_binned, recorder
+    return detector_1st, detector, extended_neuber, recorder
 
 
 def _compute_damage_and_lifetimes_RAM(assessment_parameters, recorder, component_woehler_curve_P_RAM, result):
@@ -493,17 +487,13 @@ def _compute_hcm_RAJ(assessment_parameters, scaled_load_sequence, maximum_absolu
     E, K_prime, n_prime, K_p = assessment_parameters[["E", "K_prime", "n_prime", "K_p"]]
     seeger_beste = pylife.materiallaws.notch_approximation_law_seegerbeste.SeegerBeste(E, K_prime, n_prime, K_p)
 
-    # wrap the notch approximation law by a binning class, which precomputes the values
-    seeger_beste_binned = pylife.materiallaws.notch_approximation_law.NotchApproxBinner(
-        seeger_beste
-    ).initialize(maximum_absolute_load)
-
     # create recorder object
     recorder = pylife.stress.rainflow.recorders.FKMNonlinearRecorder()
 
     # create detector object
     detector = pylife.stress.rainflow.fkm_nonlinear.FKMNonlinearDetector(
-        recorder=recorder, notch_approximation_law=seeger_beste_binned, binner=None)
+        recorder=recorder, notch_approximation_law=seeger_beste
+    )
     detector_1st = copy.deepcopy(detector)
 
     # perform HCM algorithm, first run
@@ -512,7 +502,7 @@ def _compute_hcm_RAJ(assessment_parameters, scaled_load_sequence, maximum_absolu
     # perform HCM algorithm, second run
     detector.process_hcm_second(scaled_load_sequence)
 
-    return detector_1st, detector, seeger_beste_binned, recorder
+    return detector_1st, detector, seeger_beste, recorder
 
 
 def _compute_damage_and_lifetimes_RAJ(assessment_parameters, recorder, component_woehler_curve_P_RAJ, result):

diff --git a/tests/strength/test_damage_calculator.py b/tests/strength/test_damage_calculator.py
@@ -92,20 +92,12 @@ def test_woehler_curve_P_RAM_collective_has_no_index():
 
     load_sequence_list = pd.Series([143.696, -287.392, 143.696, -359.240, 287.392, 0.000, 287.392, -287.392])
 
-    # wrap the notch approximation law by a binning class, which precomputes the values
-    maximum_absolute_load = max(np.abs(load_sequence_list))
-    extended_neuber_binned = (
-        pylife.materiallaws.notch_approximation_law.NotchApproxBinner(
-            extended_neuber
-        ).initialize(maximum_absolute_load)
-    )
-
     # create recorder object
     recorder = pylife.stress.rainflow.recorders.FKMNonlinearRecorder()
 
     # create detector object
     detector = pylife.stress.rainflow.fkm_nonlinear.FKMNonlinearDetector(
-        recorder=recorder, notch_approximation_law=extended_neuber_binned, binner=None
+        recorder=recorder, notch_approximation_law=extended_neuber
     )
 
     # perform HCM algorithm, first run
@@ -202,20 +194,12 @@ def test_woehler_curve_P_RAM_collective_has_MultiIndex():
 
     load_sequence_list = pd.Series([143.696, -287.392, 143.696, -359.240, 287.392, 0.000, 287.392, -287.392])
 
-    # wrap the notch approximation law by a binning class, which precomputes the values
-    maximum_absolute_load = max(np.abs(load_sequence_list))
-    extended_neuber_binned = (
-        pylife.materiallaws.notch_approximation_law.NotchApproxBinner(
-            extended_neuber
-        ).initialize(maximum_absolute_load)
-    )
-
     # create recorder object
     recorder = pylife.stress.rainflow.recorders.FKMNonlinearRecorder()
 
     # create detector object
     detector = pylife.stress.rainflow.fkm_nonlinear.FKMNonlinearDetector(
-        recorder=recorder, notch_approximation_law=extended_neuber_binned, binner=None
+        recorder=recorder, notch_approximation_law=extended_neuber
     )
 
     # perform HCM algorithm, first run
@@ -300,19 +284,12 @@ def test_woehler_curve_P_RAJ_has_no_index():
     seeger_beste = pylife.materiallaws.notch_approximation_law_seegerbeste.SeegerBeste(E, K_prime, n_prime, K_p)
 
     load_sequence_list = pd.Series([143.696, -287.392, 143.696, -359.240, 287.392, 0.000, 287.392, -287.392])
-
-    # wrap the notch approximation law by a binning class, which precomputes the values
-    maximum_absolute_load = max(np.abs(load_sequence_list))
-    seeger_beste_binned = pylife.materiallaws.notch_approximation_law.NotchApproxBinner(
-        seeger_beste
-    ).initialize(maximum_absolute_load)
-
     # create recorder object
     recorder = pylife.stress.rainflow.recorders.FKMNonlinearRecorder()
 
     # create detector object
     detector = pylife.stress.rainflow.fkm_nonlinear.FKMNonlinearDetector(
-        recorder=recorder, notch_approximation_law=seeger_beste_binned, binner=None
+        recorder=recorder, notch_approximation_law=seeger_beste
     )
 
     # perform HCM algorithm, first run
@@ -410,18 +387,12 @@ def test_woehler_curve_P_RAJ_has_MultiIndex():
 
     load_sequence_list = pd.Series([143.696, -287.392, 143.696, -359.240, 287.392, 0.000, 287.392, -287.392])
 
-    # wrap the notch approximation law by a binning class, which precomputes the values
-    maximum_absolute_load = max(np.abs(load_sequence_list))
-    seeger_beste_binned = pylife.materiallaws.notch_approximation_law.NotchApproxBinner(
-        seeger_beste
-    ).initialize(maximum_absolute_load)
-
     # create recorder object
     recorder = pylife.stress.rainflow.recorders.FKMNonlinearRecorder()
 
     # create detector object
     detector = pylife.stress.rainflow.fkm_nonlinear.FKMNonlinearDetector(
-        recorder=recorder, notch_approximation_law=seeger_beste_binned, binner=None
+        recorder=recorder, notch_approximation_law=seeger_beste
     )
 
     # perform HCM algorithm, first run