def fit_temperature_stable(run_num,
df_info,
df_raw,
ycol,
ystd,
ystd_sf,
remove_data=True,
plotfile=None,
loss='linear'):
# loss should be 'linear' or 'huber'
# query raw data to get run
df_i = df_info.iloc[run_num]
df_ = df_raw.query(f'"{df_i.t0}" < index < "{df_i.tf}"').copy()
df_['run_hours'] = (df_.index - df_.index[0]).total_seconds() / 60**2
# if run is long enough, only fit first 100 hours
if df_['run_hours'].max() > 100.:
df_2 = df_.query('run_hours < 100')
else:
df_2 = df_
# TESTING
# df_2_ = df_2.copy()
# df_2 = df_2.iloc[120:]
# results = None; fig = None; ax1 = None; ax2 = None
# df_ = df_.iloc[120:].copy()
# df_['run_num'] = run_num
# return results, df_, fig, ax1, ax2
# END TESTING
# create an array for weights if float supplied
if type(ystd) != np.ndarray:
# ystd = ystd*np.ones(len(df_2))
ystd = ystd * df_2[ycol] # fractional stddev. from manufacturer
# setup lmfit model
# y = A + B * np.exp(- x / C)
model = lm.Model(mod_exp, independent_vars=['x'])
params = lm.Parameters()
params.add('A', value=0, vary=True)
params.add('B', value=0, vary=True)
params.add('C', value=1, min=0, vary=True)
# fit
result = model.fit(df_2[ycol].values,
x=df_2.run_hours.values,
params=params,
weights=1 / ystd,
scale_covar=False,
method='least_squares',
fit_kws={'loss': loss})
# plot
# label for fit
# label= (r'$\underline{y = A + B e^{-x / C}}$'+'\n\n'+
# rf'$A = {result.params["A"].value:0.2f}$'+'\n'+
# rf'$B = {result.params["B"].value:0.2f}$'+'\n'+
# rf'$C = {result.params["C"].value:0.2f}$'+'\n'+
# rf'$\chi^2_\mathrm{{red.}} = {result.redchi:0.2f}$'+'\n\n')
label = (r'$\underline{y = A + B e^{-x / C}}$' + '\n\n' +
rf'$A = {result.params["A"].value:0.3f}$' +
rf'$\pm{result.params["A"].stderr:0.3f}$' + '\n' +
rf'$B = {result.params["B"].value:0.3f}$' +
rf'$\pm{result.params["B"].stderr:0.3f}$' + '\n' +
rf'$C = {result.params["C"].value:0.3f}$' +
rf'$\pm{result.params["C"].stderr:0.3f}$' + '\n' +
rf'$\chi^2_\mathrm{{red.}} = {result.redchi:0.2f}$' + '\n\n')
# set up figure with two axes
config_plots()
fig = plt.figure()
ax1 = fig.add_axes((0.12, 0.33, 0.8, 0.58))
ax2 = fig.add_axes((0.12, 0.13, 0.8, 0.2))
# plot data and fit
# data
if ystd_sf == 1:
label_data = 'Data'
else:
label_data = r'Data (error $\times$' + f'{ystd_sf})'
ax1.errorbar(df_2.index.values,
df_2[ycol].values,
yerr=ystd_sf * ystd,
fmt='o',
ls='none',
ms=2,
zorder=100,
label=label_data)
# fit
ax1.plot(df_2.index.values,
result.best_fit,
linewidth=2,
color='red',
zorder=99,
label=label)
# time constant
x_ = df_2.index[0] + timedelta(hours=result.params["C"].value)
ymin = np.min(df_2[ycol]) * 0.95
ymax = np.max(df_2[ycol]) * 1.02
ax1.plot([x_, x_], [ymin, ymax],
'--',
color='gray',
zorder=101,
label=rf'$C = {result.params["C"].value:0.3f}$ [Hours]')
# calculate residual (data - fit)
res = df_2[ycol].values - result.best_fit
res_full = (df_[ycol].values -
mod_exp(df_['run_hours'].values, **result.params))
df_['stable_temp_res'] = res_full
# quick histogram for residuals
#_ = res_full
_ = df_['NMR [T]'].values
mean_full = np.mean(_)
std_full = np.std(_)
nstd = 2
fig_h, ax_h = plt.subplots()
n, bins, patches = ax_h.hist(_,
bins=25,
histtype='step',
label='Measurements')
#label='Fit Residuals')
nm = np.max(n)
ax_h.plot([mean_full - nstd * std_full, mean_full - nstd * std_full],
[0, nm],
'r--',
label=f'{nstd} RMS from mean')
ax_h.plot([mean_full + nstd * std_full, mean_full + nstd * std_full],
[0, nm], 'r--')
# ax_h.set_xlabel(r'Fit Residuals [$^\circ$C]')
ax_h.set_xlabel(r'NMR [T]')
ax_h.set_ylabel('Count')
ax_h.set_yscale('log')
ax_h.legend()
fig_h.suptitle(
f'Temperature Stability Fit Residuals: Exponential Decay\n' +
f'Run Index {run_num}, {loss.capitalize()} Loss')
if not plotfile is None:
fig_h.savefig(plotfile + '_res_hist.pdf')
fig_h.savefig(plotfile + '_res_hist.png')
# calculate ylimit for ax2
yl = 1.1 * (np.max(np.abs(res)) + ystd_sf * ystd[0])
# plot residual
# zero-line
xmin = np.min(df_2.index.values)
xmax = np.max(df_2.index.values)
ax2.plot([xmin, xmax], [0, 0], 'k--', linewidth=2, zorder=98)
# residual
ax2.errorbar(df_2.index.values,
res,
yerr=ystd_sf * ystd,
fmt='o',
ls='none',
ms=2,
zorder=99)
# formatting
# set ylimit ax2
ax2.set_ylim([-yl, yl])
# remove ticklabels for ax1 xaxis
ax1.set_xticklabels([])
# axis labels
ax2.set_xlabel('Datetime [2021 MM-DD hh:mm]')
ax2.set_ylabel(r'(Data - Fit) [$^{\circ}$C]')
ax1.set_ylabel(r'Yoke (center magnet) Temp. [$^{\circ}$C]')
# force consistent x axis range for ax1 and ax2
#tmin = np.min(df_2.index.values)
#tmax = np.max(df_2.index.values)
#ax1.set_xlim([tmin, tmax])
#ax2.set_xlim([tmin, tmax])
# turn on legend
ax1.legend().set_zorder(102)
# add title
fig.suptitle(f'Temperature Stability Fit: Exponential Decay\n' +
f'Run Index {run_num}, {loss.capitalize()} Loss')
# minor ticks
ax1.yaxis.set_minor_locator(AutoMinorLocator())
ax2.yaxis.set_minor_locator(AutoMinorLocator())
# inward ticks and ticks on right and top
ax1.tick_params(which='both',
direction='in',
top=True,
right=True,
bottom=True)
# ax1.ticklabel_format(axis='y', useOffset=True)
ax2.tick_params(which='both', direction='in', top=True, right=True)
# tick label format consistent
formatter = DateFormatter('%m-%d %H:%M')
ax2.xaxis.set_major_formatter(formatter)
# rotate label for dates
ax2.xaxis.set_tick_params(rotation=15)
# save figure
#fig.tight_layout()
if not plotfile is None:
fig.savefig(plotfile + '.pdf')
fig.savefig(plotfile + '.png')
if remove_data:
# remove first "time constant" of data
hmax = df_.run_hours.max()
tau = result.params["C"].value
# special case (water chiller failed) -- only remove 1/2 time constant
# if hmax/tau < 1:
if hmax / tau < 1.5:
# df_ = df_.query(f'run_hours > {tau/2}').copy()
df_ = df_.query(f'run_hours > {tau/8}').copy()
# else normal case -- remove one time constant
else:
df_ = df_.query(f'run_hours > {tau}').copy()
# add run number to df
df_['run_num'] = run_num
return result, df_, fig, ax1, ax2
import numpy as np
import pandas as pd
import lmfit as lm
from scipy.interpolate import interp1d
import matplotlib.pyplot as plt
from plotting import config_plots
config_plots()
# pickle file
ddir = '/home/ckampa/data/hallprobecalib_extras/datafiles/magnet_ramp/'
# FEMM data
df_f0 = pd.read_csv(ddir+'gap75_B_vs_I_r0z0_0-300_results.txt', skiprows=8, names=['I','B'], delimiter=',')
df_fp = pd.read_csv(ddir+'gap75_B_vs_I_r0z37.5_results.txt', skiprows=8, names=['I', 'B'], delimiter=',')
df_f0 = df_f0.query('I <= 140').copy()
df_f0.eval('I = I*2', inplace=True)
df_fp = df_fp.query('I <= 140').copy()
df_fp.eval('I = I*2', inplace=True)
df_f0['B_ratio'] = df_f0.B / df_fp.B
# FEMM df
df_f = pd.read_csv(ddir+'gap75_B_vs_I_r0z0_0-300_results.txt', skiprows=8, names=['I','B'], delimiter=',')
df_f = df_f.query('I <= 140').copy()
df_f.eval('I = I*2', inplace=True)
# df = pd.read_pickle('/home/ckampa/Desktop/slow_controls_current.pkl')
df = pd.read_pickle(ddir+'ramp_2021-02-24_processed.pkl')
probe = '6A0000000D61333A'
# Is_set = np.array([224., 240.]) # testing equipment
def linear_temperature_regression(run_num, df, plotfile, xcol, ycol, ystd,
ystd_sf, force_decreasing):
# query preprocessed data to get run
df_ = df.query(f'run_num == {run_num}').copy()
# create an array for weights if float supplied
if type(ystd) != np.ndarray:
ystd = ystd*np.ones(len(df_))
# setup lmfit model
# y = A + B * x
model = lm.Model(mod_lin, independent_vars=['x'])
params = lm.Parameters()
params.add('A', value=0, vary=True)
params.add('B', value=0, vary=True)
params.add('X0', value=T0, vary=False)
# fit
result = model.fit(df_[ycol].values, x=df_[xcol].values,
params=params, weights=1/ystd, scale_covar=False)
# check if not decreasing and rerun
if force_decreasing and (result.params['B'].value > 0.):
params['B'].vary = False
result = model.fit(df_[ycol].values, x=df_[xcol].values,
params=params, weights=1/ystd, scale_covar=False)
# label for fit
bv_str = f'{result.params["B"].value:0.3E}'
ind = bv_str.find('E')
b0 = bv_str[:ind]
bsf = int(bv_str[ind+1:])
be = f'{result.params["B"].stderr/10**bsf:0.3f}'
label= (rf'$\underline{{y = A + B (x - {T0})}}$'+'\n'+
rf'$A = {result.params["A"].value:0.7f}$'+'\n'+
rf'$\pm {result.params["A"].stderr:0.7f}$'+'\n'+
rf'$B =$'+'\n'+rf'$({b0}\pm{be})$'+
rf'$\times 10^{{ {bsf} }}$'+' (fixed)\n'+
# rf'$B = {result.params["B"].value:0.1f}$'+
# rf'$\pm {result.params["B"].stderr:0.1f}$'+' (fixed)\n'+
rf'$\chi^2_\mathrm{{red.}} = {result.redchi:0.2f}$'+'\n')
else:
# label for fit
bv_str = f'{result.params["B"].value:0.3E}'
ind = bv_str.find('E')
b0 = bv_str[:ind]
bsf = int(bv_str[ind+1:])
be = f'{result.params["B"].stderr/10**bsf:0.3f}'
label= (rf'$\underline{{y = A + B (x - {T0})}}$'+'\n'+
rf'$A = {result.params["A"].value:0.7f}$'+'\n'+
rf'$\pm {result.params["A"].stderr:0.7f}$'+'\n'+
rf'$B =$'+'\n'+rf'$({b0}\pm{be})$'+
rf'$\times 10^{{ {bsf} }}$'+'\n'+
# rf'$B = {result.params["B"].value:0.3E}$'+
# rf'$\pm {result.params["B"].stderr:0.3E}$'+'\n'+
rf'$\chi^2_\mathrm{{red.}} = {result.redchi:0.2f}$'+'\n')
# plot
# set up figure with two axes
config_plots()
fig = plt.figure()
ax1 = fig.add_axes((0.12, 0.3, 0.7, 0.6))
ax2 = fig.add_axes((0.12, 0.08, 0.7, 0.185))
# colorbar axis
cb_ax = fig.add_axes([0.85, 0.15, 0.05, 0.7])
# plot data and fit
# data
# with errorbars
if ystd_sf == 1:
label_data = 'Data'
else:
label_data = r'Data (error $\times$'+f'{ystd_sf})'
ax1.errorbar(df_[xcol].values, df_[ycol].values, yerr=ystd_sf*ystd,
fmt='o', ls='none', ms=2, zorder=100, label=label_data)
# scatter with color
#sc = ax1.scatter(df_[xcol].values, df_[ycol].values, c=df_.index, s=1,
# zorder=101)
sc = ax1.scatter(df_[xcol].values, df_[ycol].values, c=df_.run_hours, s=1,
zorder=101)
# fit
ax1.plot(df_[xcol].values, result.best_fit, linewidth=2, color='red',
zorder=99, label=label)
# calculate residual (data - fit)
res = df_[ycol].values - result.best_fit
# calculate ylimit for ax2
yl = 1.1*(np.max(np.abs(res)) + ystd_sf*ystd[0])
# plot residual
# zero-line
xmin = np.min(df_[xcol].values)
xmax = np.max(df_[xcol].values)
ax2.plot([xmin, xmax], [0, 0], 'k--', linewidth=2, zorder=98)
# residual
ax2.errorbar(df_[xcol].values, res, yerr=ystd_sf*ystd, fmt='o', ls='none',
ms=2, zorder=99)
#ax2.scatter(df_[xcol].values, res, c=df_.index, s=1, zorder=101)
ax2.scatter(df_[xcol].values, res, c=df_.run_hours, s=1, zorder=101)
# colorbar for ax1
cb = fig.colorbar(sc, cax=cb_ax)
cb.set_label('Time [Hours]')
## WITH DATETIME
#cb.set_t
# print(f'vmin = {sc.colorbar.vmin}')
# print(f'vmax = {sc.colorbar.vmax}')
# # change colobar ticks labels and locators
# cb.set_ticks([sc.colorbar.vmin + t*(sc.colorbar.vmax-sc.colorbar.vmin)
# for t in cb.ax.get_yticks()])
# cbtls = [mdates.datetime.datetime.fromtimestamp((sc.colorbar.vmin +
# t*(sc.colorbar.vmax-sc.colorbar.vmin))*1e-9).strftime('%c')
# for t in cb.ax.get_yticks()]
# cb.set_ticklabels(cbtls)
#cb.ax.set_yticklabels(df_.index.strftime('%m-%d %H:%M'))
# formatting
# kludge for NMR or Hall
if ycol == 'NMR [T]':
ylabel1 = r'$|B|_\mathrm{NMR}$ [T]'
title_prefix = 'NMR'
else:
ylabel1 = r'$|B|_\mathrm{Hall}$ [T]'
title_prefix = 'Hall Probe'
# set ylimit ax2
ax2.set_ylim([-yl, yl])
# remove ticklabels for ax1 xaxis
ax1.set_xticklabels([])
# axis labels
ax2.set_xlabel(r'Yoke (center magnet) Temp. [$^{\circ}$C]')
ax2.set_ylabel('(Data - Fit) [T]')
ax1.set_ylabel(ylabel1)
# force consistent x axis range for ax1 and ax2
tmin = np.min(df_[xcol].values)
tmax = np.max(df_[xcol].values)
range_t = tmax-tmin
ax1.set_xlim([tmin-0.1*range_t, tmax+0.1*range_t])
ax2.set_xlim([tmin-0.1*range_t, tmax+0.1*range_t])
# turn on legend
ax1.legend().set_zorder(102)
# add title
fig.suptitle(f'{title_prefix} vs. Temperature Regression: Linear Model\n'+
f'Run Index {run_num}')
# minor ticks
ax1.xaxis.set_minor_locator(AutoMinorLocator())
ax2.xaxis.set_minor_locator(AutoMinorLocator())
ax1.yaxis.set_minor_locator(AutoMinorLocator())
ax2.yaxis.set_minor_locator(AutoMinorLocator())
# inward ticks and ticks on right and top
ax1.tick_params(which='both', direction='in', top=True, right=True,
bottom=True)
ax2.tick_params(which='both', direction='in', top=True, right=True)
# save figure
#fig.tight_layout()
fig.savefig(plotfile+'.pdf')
fig.savefig(plotfile+'.png')
return result, fig, ax1, ax2
