Select Git revision
evaluation.py
evaluation.py 13.24 KiB
"""Contains all evaluation classes and functions.
class LinearRegressionFitting: Represents a Linear Regression Fitting of a 2D x,y plot
"""
import os
import pandas as pd
import numpy as np
import altair as alt
import datetime
# Colors
rwthcolors= {'blau' : '#00549F',
'schwarz' : '#000000',
'magenta' : '#E30066',
'gelb' : '#FFED00',
'petrol' : '#006165',
'türkis' : '#0098A1',
'grün' : '#57AB27',
'maigrün' : '#BDCD00',
'orange' : '#F6A800',
'rot' : '#CC071E',
'bordeaux' : '#A11035',
'violett' : '#612158',
'lila' : '#7A6FAC',
}
colorlist = [rwthcolors['blau'],
rwthcolors['türkis'],
rwthcolors['grün'],
rwthcolors['orange'],
rwthcolors['violett'],
]
# Class Definitions
class LinearRegressionFitting:
"""Represents a Linear Regression Fitting of a 2D x,y plot"""
def __init__(self, PandaDataFrame):
"""Initializes the instance with a PandaDataFrame
Attributes:
x
y
slope
intercept
residuals
diagnostics
model
R2
df
chart
"""
self.input = PandaDataFrame
self.x = PandaDataFrame.to_numpy()[:, 0]
self.y = PandaDataFrame.to_numpy()[:, 1]
# https://numpy.org/doc/stable/reference/generated/numpy.polyfit.html
[self.slope, self.intercept], self.residuals, *self.diagnostics = np.polyfit(self.x, self.y, 1, full = True)
# model = expression of polynom from polyfit, here basicly
# y = model(x) <=> same as writing y = slope * x + intercept
self.model = np.poly1d([self.slope, self.intercept])
# Bestimmtheitsmaß / Coefficient of determination / R²
# 1 - ( residual sum of squares / SUM( (yi - y_mean)² ) )
self.R2 = 1 - (self.residuals[0] /
np.sum( pow( self.y - np.average( self.y ), 2) )
)
self.df = pd.DataFrame({'x': self.x,'f(x)': self.model(self.x)})
self.chart = alt.Chart(self.df).mark_line().encode(x="x", y='f(x)')
class RT_Evaluation(LinearRegressionFitting):
def __init__(self, PandaDataFrame, contactlenght):
super().__init__(PandaDataFrame)
#self.input
#self.x
#self.y
#self.slope
#self.intercept
#self.residuals
#self.diagnostics
#self.model
#self.R2
#self.df
model_df_x = np.linspace(0, self.x[-1])
self.df = pd.DataFrame({'x': model_df_x,'f(x)': self.model(model_df_x)})
#self.chart
self.chart = alt.Chart(self.df).mark_line().encode(x="x", y='f(x)')
self.contactlenght = contactlenght
# Kontaktwiderstand Rc [Ohm*mm]
# = RT(d=0)/2 [Ohm] * Contactlenght [µm/1000]
self.Rc = (self.intercept/2) * (contactlenght/1000)
# Schichtwiderstand [Ohm/sq = Ohm]
# = slope [Ohm/µm] * Contactlenght [µm]
self.Rsh = self.slope * contactlenght
# Transferlänge LT [mm] RT(d) = 0
# = slope [Ohm/µm] / RT(d=0) [Ohm] / 1000
#self.LT = self.intercept / self.slope / 1000
# Transferlänge LT [µm] RT(d) = 0
# = slope [Ohm/µm] / RT(d=0) [Ohm]
self.LT = self.intercept / self.slope / 2
# LT = sqrt(rhoc/Rsh).
# "Semiconductor Material and Device Characterization Third Edition",
# D. Schroder, p. 140, Eq. 3.21
# Spezifischer Kontaktwiderstand rhoc = LT² * Rsh
# = Ohm cm² | µm²*0.00000001 = cm²
self.rhoc = self.LT*self.LT * self.Rsh * 1E-4 * 1E-4
class TlmMeasurement(object):
def __init__(self, filelist, distances = (5, 10, 15, 20, 50), contactlenght = 50):
"""filelist as tuple
distances as tuple ## Abstände der TLM in µm
contactlenght # Kontaktweite der TLM Strukturen in µm
"""
self._creation_date = datetime.datetime.now()
self.filelist = filelist
self.path, self.files = self.importfiles(filelist)
self.distances = distances
self.contactlenght = contactlenght
self.df = self.construct_dataframes(self.filelist)
#self.df_org = self.df
self.R, self.lin_reg_charts, self.R_statistics = self.R_from_lin_reg()
self.RT0 = pd.DataFrame({'d/µm':self.distances, 'R_T/Ohm':self.R})
self.eval0 = RT_Evaluation(self.RT0, self.contactlenght)
self.eval1, self.eval2 = self.find_RT1_RT2()
self.refined = False
#self.results = self.results()
def importfiles(self, filelist):
if not len(filelist) == 5:
raise Exception("Files Missing - I need 5 files for TLM or CTLM")
filenames = []
name = ""
measurement = []
for i in range(len(filelist)):
filenames.append(os.path.split(filelist[i])[1])
name, end = filenames[i].rsplit(sep="_",maxsplit=1)
#print(name, "_", end)
if i == 0:
firstname = name
elif not name == firstname:
print("Files:", filenames)
raise Exception("Filenames differ")
measurement.append(end.split(sep=".")[0])
#print(measurement)
#print(i, len(files)-1)
if (i == len(filelist)-1) and (not measurement == ['1', '2', '3', '4', '5']):
print("Files:", filenames)
raise Exception("Not Measurement _1 to _5?")
path = os.path.split(filelist[0])[0]
return (path, filenames)
def construct_dataframes(self, filelist = None):
if filelist is None:
filelist = self.filelist
df = []
for i in range(len(filelist)):
df.append(pd.read_csv(filelist[i], sep=" ", skip_blank_lines=True, header=3, index_col=0, usecols=[0,1,2,3], names=["#","U/V","I/A","R/Ohm"]))
return df
def construct_ui_charts(self, PandaDataFrames = None):
if PandaDataFrames is None:
PandaDataFrames = self.df
uicharts = []
for i in range(len(PandaDataFrames)):
uicharts.append( alt.Chart(self.df[i]).mark_point(color=colorlist[i]).encode(x="U/V", y="I/A").interactive() )
return uicharts
def R_from_lin_reg(self, PandaDataFrames = None):
if PandaDataFrames is None:
PandaDataFrames = self.df
lin_reg_results = [] # R
lin_reg_charts = []
statistics_string = f"Statistics for Fitting of R Values \n d / µm | R² | RT / Ohm"
#print(f"Statistics for Fitting of R Values \n {'d / µm' : >9} | {'R²' : <6} | RT / Ohm ")
for i in range(len(PandaDataFrames)):
fit = LinearRegressionFitting(PandaDataFrames[i])
lin_reg_results.append( 1 / (fit.slope) )
lin_reg_charts.append( alt.Chart(fit.df).mark_line(color=colorlist[i]).encode(x="x", y='f(x)') )
#print(f"{self.distances[i] : 10d} | {fit.R2 : >5.4f} | {lin_reg_results[i] : 10.4f}")
statistics_string += "\n" + f"{self.distances[i] : 10d} | {fit.R2 : >5.4f} | {lin_reg_results[i] : 10.4f}"
#print(statistics_string)
return lin_reg_results, lin_reg_charts, statistics_string
def contruct_lin_reg_fit_charts(self):
return self.lin_reg_charts
def find_RT1_RT2(self, R = None):
if R is None:
R = self.RT0
# Remove One Measurement
# Find maximal Bestimmtheitsmaß / Coefficient of determination / R²
save_RT1 = None
max = 0
for i in range(len(R)):
Ri = R.drop(labels=i, axis=0)
evali = RT_Evaluation(Ri, self.contactlenght)
if evali.R2 >= max:
max = evali.R2
save_RT1 = evali
# Remove Two Measurements
list_of_rows_to_remove = [[0,1],
[0,2],
[0,3],
[0,4],
[1,2],
[1,3],
[1,4],
[2,3],
[2,4],
[3,4],
]
save_RT2 = None
max = 0
for rows in list_of_rows_to_remove:
Ri = R.drop(labels=rows, axis=0)
evali = RT_Evaluation(Ri, self.contactlenght)
if evali.R2 >= max:
max = evali.R2
save_RT2 = evali
return save_RT1, save_RT2
def refine_evaluated_range(self, newrange = None):
if not type(newrange) is tuple:
raise Exception("Wrong format for Newrange! Should be tuple (float, float)")
if not len(newrange) == 2:
raise Exception("Two few or many values in newrange! Should be tuple (float, float)")
minI, maxI = newrange
refined_df = self.df
for i in range(len(refined_df)):
# First vector where values for I/A are bigger than min and smaller than max = True
# (dataframe.iloc[:,1] > min) & (dataframe.iloc[:,1] < max )
# then this boolean vector is used as indexer for the dame dataframe
# every value with False is sliced
work = refined_df[i]
#print(work)
#print((dataframe.iloc[:,1] > minI) & (dataframe.iloc[:,1] < maxI ))
work = work [ (work.iloc[:,1] > minI) & (work.iloc[:,1] < maxI ) ]
#print(work)
refined_df[i] = work
self.df = refined_df
self.R, self.lin_reg_charts, statistics_string = self.R_from_lin_reg()
self.RT0 = pd.DataFrame({'d/µm':self.distances, 'R_T/Ohm':self.R})
self.eval0 = RT_Evaluation(self.RT0, self.contactlenght)
self.eval1, self.eval2 = self.find_RT1_RT2()
self.refined = True
def results(self, comment=""):
#print("Path:\n", ctlm.path, "\n", sep = "")
results_string = "Path: " + self.path + "\n"
#print("Files: {}, {}, {}, {}, {} \n".format(*ctlm.files))
results_string += "Files: {}, {}, {}, {}, {} \n".format(*self.files)
#header = "Feld Rsh R² Rc LT rhoc min I max I # removed values"
#header = "Feld Rsh R² Rc LT rhoc # removed values"
results_string += comment + " \n"
results_string += "Feld Rsh R² Rc LT rhoc # removed values \n"
#units = "- [Ohm/sq] - [Ohm mm] [µm] [Ohm cm²] [A] [A] -"
#units = "- [Ohm/sq] - [Ohm mm] [µm] [Ohm cm²] -"
results_string += "- [Ohm/sq] - [Ohm mm] [µm] [Ohm cm²] - \n"
format_string = "{:<4}{:7.2f} {:8.4f} {:6.2f} {:8.2f} {:8.2e} {} \n"
results_string += format_string.format(self.files[0].rsplit("_")[0],
self.eval0.Rsh,
self.eval0.R2,
self.eval0.Rc,
self.eval0.LT,
self.eval0.rhoc,
#minI,
#maxI,
None,
)
results_string += format_string.format(self.files[0].rsplit("_")[0],
self.eval1.Rsh,
self.eval1.R2,
self.eval1.Rc,
self.eval1.LT,
self.eval1.rhoc,
#minI,
#maxI,
1,
)
results_string += format_string.format(self.files[0].rsplit("_")[0],
self.eval2.Rsh,
self.eval2.R2,
self.eval2.Rc,
self.eval2.LT,
self.eval2.rhoc,
#minI,
#maxI,
2,
)
return results_string
def construct_rt_charts(self):
rt_charts = []
rt_charts.append( alt.Chart(self.RT0).mark_point(color=rwthcolors['blau']).encode(x='d/µm', y='R_T/Ohm').interactive() )
rt_charts.append( alt.Chart(self.eval0.df).mark_line(color=rwthcolors['bordeaux']).encode(x='x', y='f(x)') )
rt_charts.append( alt.Chart(self.eval1.df).mark_line(color=rwthcolors['violett']).encode(x='x', y='f(x)') )
rt_charts.append( alt.Chart(self.eval2.df).mark_line(color=rwthcolors['lila']).encode(x='x', y='f(x)') )
return rt_charts