def plot_progress(tech, T):
P_tech = T[T['tech'] == tech]['progress']
P_tech = [P for P in P_tech if len(P.shape) > 0]
nDists = np.concatenate([P['new_dist'].astype(float) for P in P_tech])
oDists = np.concatenate([P['old_dist'].astype(float) for P in P_tech])
dDists = oDists - nDists
fig, ax = plotting.subplots(1,
2,
figsize_adjust=(1.0, 0.5),
constrained_layout=True)
ax[0].hist(nDists, bins=args.hist_bins, **tech_style(tech))
ax[0].axvline(x=0, lw=1, color='black')
ax[0].set_ylabel(r'\#steps where new distance is $d$')
ax[0].set_xlabel(r'Distance ($d$ — ' + tech + r')')
ax[1].hist(dDists, bins=args.hist_bins, **tech_style(tech))
ax[1].axvline(x=0, lw=1, color='black')
ax[1].set_ylabel(r'\#steps where progress is $\delta$')
ax[1].set_xlabel(r'Progress ($\delta$ — ' + tech + r')')
plotting.show(fig,
outdir=outdir,
basefilename=filename(tech + '-dist-n-progress'),
w_pad=0.06)
def freeyaw():
"""
Selected time series of free yaw response cases
"""
fpath = 'data/calibrated/DataFrame'
fdir = 'figures/freeyaw'
fname1 = '0405_run_277_9.0ms_dc1_flexies_freeyaw_highrpm.h5'
fname2 = '0410_run_330_9ms_dc1_samoerai_freeyaw_highrpm.h5'
flex = pd.read_hdf(os.path.join(fpath, fname1), 'table')
samo = pd.read_hdf(os.path.join(fpath, fname2), 'table')
fname = '277-vs-330-9ms-rpm'
fig, axes = plotting.subplots(nrows=1, ncols=1, figsize=(5,2), dpi=120)
axes = axes.flatten()
duration = 15
ax = axes[0]
t0 = 48
i0 = int(t0*129024/63.0)
i1 = i0 + int(duration*129024/63.0)
ax.plot(flex.time[i0:i1]-t0, flex.rpm[i0:i1], 'b-', label='straight blade')
t0 = 49.6
i0 = int(t0*69000/69.0)
i1 = i0 + int(duration*69000/69.)
ax.plot(samo.time[i0:i1]-t0, samo.rpm[i0:i1], 'r-', label='swept blade', alpha=0.7)
for ax in axes:
leg = ax.legend(loc='best', borderaxespad=0)
leg.get_frame().set_alpha(0.6)
ax.grid(True)
ax.set_xlabel('time [s]')
ax.set_ylabel('RPM')
ax.set_xlim([0, 15])
fig.tight_layout()
fig.subplots_adjust(top=0.92)
# fig.suptitle('windspeed 9 m/s')
print('saving: %s' % os.path.join(fdir, fname))
fig.savefig(os.path.join(fdir, fname + '.png'))
fig.savefig(os.path.join(fdir, fname + '.eps'))
# =========================================================================
fname = '277-vs-330-9ms-yaw'
fig, axes = plotting.subplots(nrows=1, ncols=1, figsize=(5,2), dpi=120)
axes = axes.flatten()
duration = 15
ax = axes[0]
t0 = 48
i0 = int(t0*129024/63.0)
i1 = i0 + int(duration*129024/63.0)
ax.plot(flex.time[i0:i1]-t0, flex.yaw_angle[i0:i1], 'b-',
label='straight blade')
t0 = 49.6
i0 = int(t0*69000/69.0)
i1 = i0 + int(duration*69000/69.)
ax.plot(samo.time[i0:i1]-t0, samo.yaw_angle[i0:i1], 'r-',
label='swept blade', alpha=0.7)
for ax in axes:
leg = ax.legend(loc='best', borderaxespad=0)
leg.get_frame().set_alpha(0.6)
ax.grid(True)
ax.set_xlabel('time [s]')
ax.set_ylabel('yaw angle [deg]')
ax.set_xlim([0, 15])
fig.tight_layout()
fig.subplots_adjust(top=0.92)
# fig.suptitle('windspeed 9 m/s')
print('saving: %s' % os.path.join(fdir, fname))
fig.savefig(os.path.join(fdir, fname + '.png'))
fig.savefig(os.path.join(fdir, fname + '.eps'))
def plot_windbluepower_st_540():
"""Plot the data and add some other constant resistance values to it.
"""
q_rmax, p_rmax, rpm_rmax, emax = torque_power_at_ohm(28)
q_rmin, p_rmin, rpm_rmin, emin = torque_power_at_ohm(11)
# headers:
# Run #,W [rpm],Rsetting [ohm],Voltage [V],Current [A],tau_input [N-m]
# P_output [W],Eff [-]
fname = "data/model/generator-windbluepower-st-540.csv"
df = pd.read_csv(fname, sep=",")
rad = df["W [rpm]"].values * np.pi / 30.0
rpm = df["W [rpm]"].values
Qc = df["tau_input [N-m]"]
R = df["Rsetting [ohm]"]
eff = df["Eff [-]"] * 100.0
# the torque-rpm grid
xi = np.linspace(rpm.min(), rpm.max(), 200)
yi = np.linspace(Qc.min(), Qc.max(), 200)
cmap = mpl.cm.get_cmap("jet", 8)
# -------------------------------------------------------------------------
# iso-lines: dump load resistance value in Ohm
# grid the data
zi = mpl.mlab.griddata(rpm, Qc, R, xi, yi, interp="linear")
# consider switching to: matplotlib.tri.Triangulation or
# matplotlib.tri.TriInterpolator, see: matplotlib.org/api/tri_api.html
fig, axes = plotting.subplots(nrows=1, ncols=1, figsize=(4, 2), dpi=120)
ax = axes.flatten()[0]
cnt = ax.contour(xi, yi, zi, 8, colors=None, cmap=cmap) # , vmax=35, vmin=0)
ax.clabel(cnt, inline=1, fontsize=10, fmt="%1.0f")
ax.set_xlabel("rotor speed [rpm]")
ax.set_ylabel("input torque [Nm]")
ax.grid()
# clb = fig.colorbar(cnt)
# clb.set_label('color bar label')
# leg = ax.legend(loc='best', labelspacing=0, columnspacing=0)
# leg.get_frame().set_alpha(0.6)
fig.tight_layout()
fig.savefig("figures/generator-st-540-contour.png")
fig.savefig("figures/generator-st-540-contour.eps")
# -------------------------------------------------------------------------
# -------------------------------------------------------------------------
# iso-lines: efficiency
# grid the data
zi = mpl.mlab.griddata(rpm, Qc, eff, xi, yi, interp="linear")
# consider switching to: matplotlib.tri.Triangulation or
# matplotlib.tri.TriInterpolator, see: matplotlib.org/api/tri_api.html
fig, axes = plotting.subplots(nrows=1, ncols=1, figsize=(4, 2), dpi=120)
ax = axes.flatten()[0]
cnt = ax.contour(xi, yi, zi, 8, colors=None, cmap=cmap) # , vmax=35, vmin=0)
ax.clabel(cnt, inline=1, fontsize=10, fmt="%1.0f")
ax.set_xlabel("rotor speed [rpm]")
ax.set_ylabel("input torque [Nm]")
ax.grid()
fig.tight_layout()
fig.savefig("figures/generator-st-540-contour-eff.png")
fig.savefig("figures/generator-st-540-contour-eff.eps")
# -------------------------------------------------------------------------
# -------------------------------------------------------------------------
fig, axes = plotting.subplots(nrows=1, ncols=1, figsize=(5, 3), dpi=120)
ax = axes.flatten()[0]
# plt.figure()
# plt.contour(xi, yi, zi,6, colors='k') #, vmax=35, vmin=0)
N = len(range(5, 30, 1))
# select a color map
cmap = mpl.cm.get_cmap("jet", N) # hot
# convert to array
cmap_arr = cmap(np.arange(N))
# and now you have each color as an RGB tuple as
# for i in cmap_arr:
# coltuple = tuple(i[0:3])
for r, color in zip(range(5, 30, 1), cmap_arr):
c = tuple(color[0:3])
q_, p_, rpm_, eff_ = torque_power_at_ohm(r)
ax.plot(rpm_, q_, ls="-", alpha=0.6, color=c, marker="+")
# plt.plot(rpm_rmax, q_rmax, 'r-', label='R=28 (dc0)')
# plt.plot(rpm_rmin, q_rmin, 'b-', label='R=11 (dc1)')
# plt.legend(loc='best')
ax.set_xlabel("rotor speed [rpm]")
ax.set_xlabel("input torque [Nm]")
ax.grid()
# leg = ax.legend(loc='best', labelspacing=0, columnspacing=0)
# leg.get_frame().set_alpha(0.6)
fig.tight_layout()
fig.savefig("figures/generator-st-540.png")
fig.savefig("figures/generator-st-540.eps")
def plots(args):
T = read_reports(args.dir)
outdir = OutputDir(args.outputs, log=True)
filename = lambda f: args.prefix + '-' + f if args.prefix is not None else f
T = T[T['crit'] == args.criterion]
T_init_tests = {
n: T[T['init_tests'] == n]
for n in np.unique(T['init_tests'])
}
for init_tests in T_init_tests:
generated_tests = sum(run['report']['#tests'][-1] - init_tests
for run in T_init_tests[init_tests])
n_runs = len(T_init_tests[init_tests])
print(f'{generated_tests} tests generated for |X_0|={init_tests}'
'(average = {} test{}/run).'.format(*s_(generated_tests * 1. /
n_runs)))
def tech_style(tech):
return dict(color='blue' if tech == 'pca' else 'red')
# Progress/ICA
def plot_progress(tech, T):
P_tech = T[T['tech'] == tech]['progress']
P_tech = [P for P in P_tech if len(P.shape) > 0]
nDists = np.concatenate([P['new_dist'].astype(float) for P in P_tech])
oDists = np.concatenate([P['old_dist'].astype(float) for P in P_tech])
dDists = oDists - nDists
fig, ax = plotting.subplots(1,
2,
figsize_adjust=(1.0, 0.5),
constrained_layout=True)
ax[0].hist(nDists, bins=args.hist_bins, **tech_style(tech))
ax[0].axvline(x=0, lw=1, color='black')
ax[0].set_ylabel(r'\#steps where new distance is $d$')
ax[0].set_xlabel(r'Distance ($d$ — ' + tech + r')')
ax[1].hist(dDists, bins=args.hist_bins, **tech_style(tech))
ax[1].axvline(x=0, lw=1, color='black')
ax[1].set_ylabel(r'\#steps where progress is $\delta$')
ax[1].set_xlabel(r'Progress ($\delta$ — ' + tech + r')')
plotting.show(fig,
outdir=outdir,
basefilename=filename(tech + '-dist-n-progress'),
w_pad=0.06)
if not args.no_pca_progress:
plot_progress('pca', T)
# Progress/ICA
if not args.no_ica_progress:
plot_progress('ica', T)
# Summary
if not args.no_summary:
def plot_style(report):
return tech_style(report['tech'])
def it_(ax):
return ax if len(T_init_tests) > 1 else [ax]
Nms = args.dnn_name # r'\mathcal{N}_{\mathsf{ms}}'
cov_label_ = lambda d, n, x: r'\mathrm{' + d + r'}(\mathcal{B}_{' + n + r', ' + x + '})'
cov_label = lambda n, x: \
cov_label_ ('BFCov', n, x) if args.criterion == 'bfc' else \
cov_label_ ('BFdCov', n, x)
fig, ax = plotting.subplots(3,
len(T_init_tests),
sharex='col',
sharey='row',
constrained_layout=True)
for axi in it_(ax[-1]):
# unshare x axes for the bottom row:
g = axi.get_shared_x_axes()
g.remove(axi)
for a in g.get_siblings(axi):
g.remove(a)
for init_tests, axi in zip(T_init_tests, it_(ax[0])):
for run in T_init_tests[init_tests]:
axi.plot(run['report']['#tests'] - init_tests,
**plot_style(run))
from matplotlib.ticker import StrMethodFormatter
for init_tests, axi in zip(T_init_tests, it_(ax[1])):
for run in T_init_tests[init_tests]:
if len(run['report']) == 0:
continue
axi.plot(
run['report']['coverage'] - run['report']['coverage'][0],
**plot_style(run))
axi.yaxis.set_major_formatter(StrMethodFormatter('{x:2.1f}'))
axi.yaxis.set_ticks(
np.arange(0, np.amax(axi.get_yticks()), step=0.1))
for init_tests, axi in zip(T_init_tests, it_(ax[2])):
init_covs = [
run['report']['coverage'][0]
for run in T_init_tests[init_tests] if len(run['report']) > 0
]
final_covs = [
run['report']['coverage'][-1]
for run in T_init_tests[init_tests] if len(run['report']) > 0
]
bp = axi.boxplot(
[init_covs, final_covs],
positions=[0, 20],
widths=6,
# labels = [r'initial ($i=0$)', 'final'],
flierprops=dict(marker='.', markersize=1),
bootstrap=1000,
manage_ticks=False)
axi.yaxis.set_major_formatter(StrMethodFormatter('{x:2.1f}'))
for box in bp['boxes']:
box.set(linewidth=.5)
for box in bp['caps']:
box.set(linewidth=.5)
plt.setp(axi.get_xticklabels(), visible=False)
for init_tests, axi in zip(T_init_tests, it_(ax[1])):
axi.xaxis.set_tick_params(which='both', labelbottom=True)
# Set labels and column titles:
for init_tests, axi in zip(T_init_tests, it_(ax[0])):
axi.set_title(f'$|X_0| = {init_tests}$')
for axi in it_(ax[-1]):
axi.set_xlabel(r'iteration ($i$)')
it_(ax[0])[0].set_ylabel(r'$|X_i| - |X_0|$')
it_(ax[1])[0].set_ylabel(r'$' + cov_label(Nms, r'X_i') + '-' +
cov_label(Nms, r'X_0') + '$')
it_(ax[2])[0].set_ylabel(r'$' + cov_label(Nms, r'X_i') + '$')
# it_(ax[-1])[(len (T_init_tests) - 1) // 2 + 1].set_xlabel (r'iteration ($i$)')
plotting.show(fig,
basefilename=filename('summary-per-X0'),
outdir=outdir,
rect=(.01, 0, 1, 1))
def get_steps(self, sigs, weights, cutoff_hz=1.0, order=2, window=4.0,
x_threshold=2.0, figname=None, min_step_window=2.0):
"""
Parameters
----------
sigs : list(ndarray(n), ndarray(n))
List of signals (1D-arrays)
"""
x = sigs[0]
y = sigs[1]
xf, xf_ds, xf_ds_dt, x_regress = self.conditioning(x, cutoff_hz=cutoff_hz,
order=order, window=window)
xf_ds_regress = x_regress[:,0]
yf, yf_ds, yf_ds_dt, y_regress = self.conditioning(y, cutoff_hz=cutoff_hz,
order=order, window=window)
yf_ds_regress = y_regress[:,0]
# select on the product of yf_ds_regress * xf_ds_regress,
# both need to be steady!
xw = weights[0]
yw = weights[1]
xy_ds = np.abs(xf_ds_regress*xw) + np.abs(yf_ds_regress*yw)
xf_sel_mask, xf_sel_arg = self.select(xy_ds, x_threshold)
step_ds_mask, steps_ds = self.steady_steps(xf_sel_mask,
step_lenght=min_step_window)
# save steps in high-res sampling of the original signal
steps = np.round(steps_ds * self.sps / self.freq_ds, 0).astype(np.int)
np.savetxt(figname.replace('.png', '_steps.txt'), steps)
# steps_ds_times = self.t_ds[steps_ds.flatten()]
# steps = np.ndarray(steps_ds.shape) * np.nan
# for k in range(steps.shape[0]):
# t0 = self.t_ds[steps[k,0]]
# t1 = self.t_ds[steps[k,1]]
# steps[k,0] = np.abs(self.time - t0).argmin()
# steps[k,0] = np.abs(self.time - t1).argmin()
if figname is not None:
print('start plotting...')
fig, axes = plotting.subplots(nrows=3, ncols=1, figsize=(8,9),
dpi=120)
ax = axes[0,0]
ax.set_title('original and filtered signals')
ax.plot(self.time, x, 'r-', alpha=0.3)
ax.plot(self.time, xf, 'r-')
ax.grid()
axr = ax.twinx()
axr.plot(self.time, y, 'g-', alpha=0.3)
axr.plot(self.time, yf, 'g-')
ax = axes[1,0]
ax.set_title('lin regr window: %1.02f sec' % window)
t_mask = self.t_ds.copy()
t_mask[~xf_sel_mask] = np.nan
x_mask = xf_ds.copy()
x_mask[~xf_sel_mask] = np.nan
ax.plot(self.t_ds, xf_ds, 'r-', alpha=1.0, label='xf ds')
ax.plot(self.t_ds, x_mask, 'k-+', alpha=0.7, label='xf select')
ax.grid()
axr = ax.twinx()
axr.plot(self.t_ds, yf_ds, 'g-', alpha=0.8, label='yx ds')
y_mask = yf_ds.copy()
y_mask[~xf_sel_mask] = np.nan
axr.plot(self.t_ds, y_mask, 'k-+', alpha=0.7, label='yf select')
xmin = axr.get_ylim()[0]
xmax = axr.get_ylim()[1]
collection = region.span_where(self.t_ds, ymin=xmin, ymax=xmax,
where=xf_sel_mask, facecolor='grey',
alpha=0.4)
axr.add_collection(collection)
leg = plotting.one_legend(ax, axr, loc='best')
leg.get_frame().set_alpha(0.5)
ax = axes[2,0]
rpl = (x_threshold, min_step_window)
ax.set_title('threshold: %1.02f, min step window: %1.2f sec' % rpl)
ax.plot(self.t_ds, np.abs(xf_ds_regress), 'r-',
label='xf lin regress', alpha=0.9)
ax.plot(self.t_ds, np.abs(yf_ds_regress), 'g-',
label='yf lin regress', alpha=0.9)
ax.plot(self.t_ds, np.abs(xy_ds), 'k-', label='xy*w', alpha=0.7)
ax.axhline(y=x_threshold, linewidth=1, color='k', linestyle='--',
aa=False)
ax.set_ylim([0,5])
xmin = ax.get_ylim()[0]
xmax = ax.get_ylim()[1]
collection = region.span_where(self.t_ds, ymin=xmin, ymax=xmax,
where=step_ds_mask, facecolor='grey',
alpha=0.4)
ax.add_collection(collection)
# axr = ax.twinx()
# axr.plot(self.t_ds, np.abs(yf_ds_regress), 'g-',
# label='yf lin regress', alpha=0.9)
# ax, axr = plotting.match_yticks(ax, axr)
# axr.set_ylim([0,5])
# leg = plotting.one_legend(ax, axr, loc='best')
# leg.get_frame().set_alpha(0.5)
ax.grid()
leg = ax.legend(loc='best')
leg.get_frame().set_alpha(0.5)
fig.tight_layout()
fig.savefig(figname)
print(figname)
return steps
