def test_one_cycle_down(self):
class_count = 4
x = np.array([4, 2, 3, 1])
x_max = 4
x_min = 1
class_width, \
class_offset = self.class_param(x, class_count)
hysteresis = class_width * 0.99
enforce_margin = 0 # First and last data point may be excluded in tp
use_HCM = 0 # Use 4 point method, not HCM
use_ASTM = 0 # Use 4 point method, not ASTM
residual_method = 0 # No processing on residue
spread_damage = 0 # No damage spreading
res = rfcnt.rfc(x,
class_count=class_count,
class_width=class_width,
class_offset=class_offset,
hysteresis=hysteresis,
residual_method=residual_method,
enforce_margin=enforce_margin,
use_HCM=use_HCM,
use_ASTM=use_ASTM,
spread_damage=spread_damage)
self.assertEqual(res["rfm"].sum(), 1)
self.assertEqual(res["rfm"][2 - 1, 3 - 1], 1)
self.assertTrue((res["res"].flatten() == [4, 1]).all())
def test_empty_series(self):
class_count = 100
x = np.array([])
x_max = 1
x_min = -1
class_width, \
class_offset = self.class_param(x, class_count)
hysteresis = class_width
enforce_margin = 0 # First and last data point may be excluded in tp
use_HCM = 0 # Use 4 point method, not HCM
use_ASTM = 0 # Use 4 point method, not ASTM
residual_method = 0 # No processing on residue
spread_damage = 0 # No damage spreading
res = rfcnt.rfc(x,
class_count=class_count,
class_width=class_width,
class_offset=class_offset,
hysteresis=hysteresis,
residual_method=residual_method,
enforce_margin=enforce_margin,
use_HCM=use_HCM,
use_ASTM=use_ASTM,
spread_damage=spread_damage)
self.assertEqual(res["rfm"].sum(), 0)
self.assertEqual(len(res["res"]), 0)
def test_small_sample(self):
class_count = 6
x = np.array([2, 5, 3, 6, 2, 4, 1, 6, 1, 4, 1, 5, 3, 6, 3, 6, 1, 5, 2])
x_max = x.max()
x_min = x.min()
class_width, \
class_offset = self.class_param(x, class_count)
hysteresis = class_width
enforce_margin = 0 # First and last data point may be excluded in tp
use_HCM = 0 # Use 4 point method, not HCM
use_ASTM = 0 # Use 4 point method, not ASTM
residual_method = 0 # No processing on residue
spread_damage = 0 # No damage spreading
res = rfcnt.rfc(x,
class_count=class_count,
class_width=class_width,
class_offset=class_offset,
hysteresis=hysteresis,
residual_method=residual_method,
enforce_margin=enforce_margin,
use_HCM=use_HCM,
use_ASTM=use_ASTM,
spread_damage=spread_damage)
self.assertEqual(res["rfm"].sum(), 7)
self.assertEqual(res["rfm"][5 - 1, 3 - 1], 2)
self.assertEqual(res["rfm"][6 - 1, 3 - 1], 1)
self.assertEqual(res["rfm"][1 - 1, 4 - 1], 1)
self.assertEqual(res["rfm"][2 - 1, 4 - 1], 1)
self.assertEqual(res["rfm"][1 - 1, 6 - 1], 2)
self.assertTrue((res["res"].flatten() == [2, 6, 1, 5, 2]).all())
def test_long_series(self):
try:
import pandas as pd
except ImportError as err:
print("This test requires module 'pandas'!")
raise err
class_count = 100
class_offset = -2025
class_width = 50
x = pd.read_csv(os.path.join(self.get_script_path(),
"long_series.csv"),
header=None)
x = x.to_numpy().squeeze()
hysteresis = class_width
use_HCM = 0
enforce_margin = 1 # Enforce first and last data point included in tp
use_HCM = 0 # Use 4 point method, not HCM
use_ASTM = 0 # Use 4 point method, not ASTM
residual_method = 0 # 0=RFC_RES_NONE, 7=RFC_RES_REPEATED
spread_damage = 1 # 0=RFC_SD_HALF_23, 1=RFC_SD_RAMP_AMPLITUDE_23
res = rfcnt.rfc(x,
class_count=class_count,
class_width=class_width,
class_offset=class_offset,
hysteresis=hysteresis,
residual_method=residual_method,
enforce_margin=enforce_margin,
use_HCM=use_HCM,
use_ASTM=use_ASTM,
spread_damage=spread_damage)
# With residuum: pd == 9.8934e-06 (repeated)
# Without residuum: pd == 1.1486e-07
self.assertTrue(
np.absolute(res["tp"][:, 2].sum() / res["damage"] - 1) < 1e-10)
spread_damage = 8
# 7=RFC_SD_TRANSIENT_23, 8=RFC_SD_TRANSIENT_23c
res = rfcnt.rfc(x,
class_count=class_count,
class_width=class_width,
class_offset=class_offset,
hysteresis=hysteresis,
residual_method=residual_method,
enforce_margin=enforce_margin,
use_HCM=use_HCM,
use_ASTM=use_ASTM,
spread_damage=spread_damage)
#assert(abs(sum(dh) / pd - 1) < 1e-10)
self.assertEqual("%.4e" % res["damage"], "1.1486e-07")
self.assertEqual(res["rfm"].sum(), 640)
self.assertEqual(len(res["res"]), 10)
test = np.absolute(
res["res"].flatten() -
[0, 142, -609, 2950, -2000, 2159, 1894, 2101, 1991, 2061])
self.assertTrue(test.sum() < 1e-3)
def example_1():
try:
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.gridspec import GridSpec
import seaborn as sns
except ImportError as err:
print(
"This example requires modules 'pandas', 'matplotlib' and 'seaborn!'"
)
raise err
data = pd.read_csv(os.path.join(__get_script_path(), "long_series.csv"),
header=None)
data = data.to_numpy().squeeze()
class_count = 50
class_range = data.max() - data.min()
class_width = class_range / (class_count - 1)
class_offset = data.min() - class_width / 2
res = rfcnt.rfc(
data,
class_count=class_count,
class_offset=class_offset,
class_width=class_width,
hysteresis=class_width,
use_HCM=0,
use_ASTM=0,
#spread_damage=0, # RFC_SD_HALF_23
spread_damage=8, # RFC_SD_TRANSIENT_23c
#residual_method=0, # RFC_RES_NONE
residual_method=7, # RFC_RES_REPEATED
wl={
"sd": 1e3,
"nd": 1e7,
"k": 5
})
fig = plt.figure(figsize=(14, 10))
gs = GridSpec(nrows=3, ncols=2, width_ratios=[1, 2])
ax1 = fig.add_subplot(gs[0, 0])
sns.heatmap(res["rfm"], cmap="YlOrRd", ax=ax1)
ax1.invert_yaxis()
plt.grid(which="both")
plt.xlabel("Class # (to)")
plt.ylabel("Class # (from)")
r = utils.rpplot_prepare(sa=res["rp"][:, 0] / 2, counts=res["rp"][:, 1])
ax2 = fig.add_subplot(gs[0, 1])
sns.lineplot(x=r["counts"].cumsum(),
y=r["sa"],
drawstyle='steps-pre',
ci=None,
ax=ax2)
plt.xscale("log")
plt.ylim(bottom=0, top=2000)
plt.xlim(left=0.9)
plt.grid(which="both")
plt.xlabel("N (log) [1]")
plt.ylabel("$S_a$")
ax3 = fig.add_subplot(gs[1, :])
sns.lineplot(x=np.arange(len(res["dh"])), y=res["dh"].cumsum(), ax=ax3)
plt.grid(which="both")
plt.xlabel("Sample #")
plt.ylabel("Damage (cumulative)")
ax4 = fig.add_subplot(gs[2, :])
sns.lineplot(x=np.arange(len(data)), y=data, ax=ax4)
plt.grid(which="both")
plt.xlabel("Sample #")
plt.ylabel("Value")
fig.tight_layout()
plt.show()