Make HCM rainflow counter agnostic of the assessment point index

src/pylife/stress/rainflow/fkm_nonlinear.py

@@ -543,17 +543,17 @@ def _adjust_samples_and_flush_for_hcm_first_run(self, samples):
         if not is_multi_index:
             samples = np.concatenate([[0], np.asarray(samples)])
         else:
-            # get the index with all node_id`s
-            node_id_index = samples.groupby("node_id").first().index
+            assessment_levels = [name for name in samples.index.names if name != "load_step"]
+            assessment_idx = samples.groupby(assessment_levels).first().index
 
             # create a new sample with 0 load for all nodes
-            multi_index = pd.MultiIndex.from_product([[0], node_id_index], names=["load_step","node_id"])
+            multi_index = pd.MultiIndex.from_product([[0], assessment_idx], names=samples.index.names)
             first_sample = pd.Series(0, index=multi_index)
 
             # increase the load_step index value by one for all samples
             samples_without_index = samples.reset_index()
             samples_without_index.load_step += 1
-            samples = samples_without_index.set_index(["load_step", "node_id"])[0]
+            samples = samples_without_index.set_index(samples.index.names)[0]
 
             # prepend the new zero load sample
             samples = pd.concat([first_sample, samples], axis=0)
@@ -568,9 +568,7 @@ def _adjust_samples_and_flush_for_hcm_first_run(self, samples):
         scalar_samples_twice = np.concatenate([scalar_samples, scalar_samples])
         turn_indices, _ = RFG.find_turns(scalar_samples_twice)
 
-        flush = True
-        if len(scalar_samples)-1 not in turn_indices:
-            flush = False
+        flush = len(scalar_samples)-1 in turn_indices
 
         return samples, flush
 

src/pylife/stress/rainflow/recorders.py

@@ -288,11 +288,12 @@ def collective(self):
         """
 
         if len(self.S_min) > 0 and len(self.S_min.index.names) > 1:
-            n_nodes = self.S_min.groupby('node_id').first().count()
-            n_hystereses = int(len(self.S_min) / n_nodes)
+            assessment_levels = [name for name in self.S_min.index.names if name != "load_step"]
+            n_assessment_points = self.S_min.groupby(assessment_levels).first().count()
+            n_hystereses = int(len(self.S_min) / n_assessment_points)
 
             index = pd.MultiIndex.from_product(
-                [range(n_hystereses), range(n_nodes)],
+                [range(n_hystereses), range(n_assessment_points)],
                 names=["hysteresis_index", "assessment_point_index"],
             )
         else:

tests/stress/rainflow/test_fkm_nonlinear.py

@@ -817,6 +817,31 @@ def test_hcm_first_second(vals, num):
     assert multi_collective.to_json(indent=4) == reference
 
 
+def test_element_id():
+    vals = np.array([100., -200., 100., -250., 200., 0., 200., -200.]) * (250+6.6)/250 * 1.4
+    signal = pd.DataFrame({11: vals, 12: 2*vals, 13: 3*vals, 'load_step': range(len(vals))}).set_index('load_step').stack()
+    signal.index.names = ['load_step', 'element_id']
+
+    E = 206e3    # [MPa] Young's modulus
+    K = 3.1148*(1251)**0.897 / (( np.min([0.338, 1033.*1251.**(-1.235)]) )**0.187)
+    #K = 2650.5   # [MPa]
+    n = 0.187    # [-]
+    K_p = 3.5    # [-] (de: Traglastformzahl) K_p = F_plastic / F_yield (3.1.1)
+
+    extended_neuber = NAL.ExtendedNeuber(E, K, n, K_p)
+
+    detector = FKMNonlinearDetector(
+        recorder=RFR.FKMNonlinearRecorder(), notch_approximation_law=extended_neuber
+    )
+    detector.process_hcm_first(signal).process_hcm_second(signal)
+
+    with open(f'tests/stress/rainflow/reference-fkm-nonlinear/reference_first_second-7.json') as f:
+        reference = f.read()
+
+    assert detector.recorder.collective.to_json(indent=4) == reference
+    assert detector.recorder.loads_max.index.names == ["load_step", "element_id"]
+
+
 @pytest.fixture
 def detector_seeger_beste():
     E = 206e3    # [MPa] Young's modulus