Device and CSV Data

Device and CSV Data

FDS writes several csv files per case: _devc.csv for &DEVC output, _hrr.csv for the global heat release rate budget, _ctrl.csv for &CTRL states, and so on. load_csv reads any of them into a pandas data frame.

Reading

import os
import pyfdstools as fds

workingDir = os.path.join(
    os.path.dirname(fds.__file__), 'examples', 'data', 'case001.zip')
chid = 'case001'

devices = fds.load_csv(workingDir, chid, '_devc')
print(devices.columns.tolist()[:4])
# ['Time', '"U"', '"burn"', '"con"']

The suffix argument selects which file:

Suffix	File
'_devc'	device output (the default)
'_hrr'	heat release rate budget
'_ctrl'	control function states
'_steps'	timestep history
'_cpu'	per-routine timings

hrr = fds.load_csv(workingDir, chid, '_hrr')
print(hrr.columns.tolist()[:5])
# ['Time', 'HRR', 'HRR_OX', 'Q_RADI', 'Q_CONV']

Column names come straight from FDS, so device columns keep the quotes FDS writes around them. Strip them if you prefer:

devices.columns = [c.strip('"') for c in devices.columns]

Why not pandas.read_csv?

FDS csv files have two header rows — units then labels — and they contain non-ASCII characters in the unit row (°C, kW/m²). Some outputs also have a trailing comma on every line. load_csv detects the header length, repairs the encoding and drops the trailing field, which pd.read_csv on its own does not.

If a file has an unusual header, pin it explicitly:

devices = fds.load_csv(workingDir, chid, '_devc', labelRow=1)

skipcols drops columns by index, which is useful when a file has a malformed one:

devices = fds.load_csv(workingDir, chid, '_devc', skipcols=[5])

Plotting a time history

import matplotlib.pyplot as plt

hrr = fds.load_csv(workingDir, chid, '_hrr')

fig, ax = plt.subplots(figsize=(8, 5))
ax.plot(hrr['Time'], hrr['HRR'], linewidth=2)
ax.set_xlabel('Time (s)')
ax.set_ylabel('HRR (kW)')
fig.tight_layout()
fig.savefig('hrr.png')

Selecting a group of devices

Devices are usually named systematically, so selecting a group is a string match on the column names:

thermocouples = [c for c in devices.columns if 'TC' in c]
peak = devices[thermocouples].max().max()

maxValuePlot and maxValueCSV plot and write a set of such groups:

import numpy as np

times = devices['Time'].values
values = devices[thermocouples].values
names = [c.strip('"') for c in thermocouples]

fds.maxValuePlot(times, values, names, 'thermocouples.png',
                 vName='Temperature (C)')
fds.maxValueCSV(times, values, names, 'thermocouples')

Smoothing a noisy trace

Device output from a turbulent simulation is noisy. Two options:

# Kalman filter: no phase lag, tune with Q (process) and R (measurement)
smoothed = fds.kalmanFilter(devices['"TC-1"'].values, Q=1e-5, R=0.25)

# Boxcar average over a window, via timeAverage
values = devices[thermocouples].values.T[:, None, :]   # (N, 1, NT)
averaged, outTimes = fds.timeAverage(values, times, window=10.0)

See Time Averaging.

Two-zone reduction

A vertical array of thermocouples can be reduced to the equivalent two-layer model from the FDS verification guide:

import numpy as np

elevations = np.array([0.5, 1.0, 1.5, 2.0, 2.5])
profile = devices[['"TC-%0.1f"' % z for z in elevations]].values[-1, :]

lower, upper, interface = fds.getTwoZone(elevations, profile)
print('lower layer %.1f C, upper layer %.1f C, interface at %.2f m'
      % (lower, upper, interface))

The elevations may be given in either direction; the ordering is detected from the data.

Building a slice from a grid of devices

If you laid out devices on a regular grid, they can be interpolated onto a slice plane and written out as a slice file smokeview will display. pyfdstools/examples/make_slice_from_devc.py does exactly this for a radiometer array, and pyfdstools.examples.exampleAddGasPhaseHeatFluxSlice is the same code as a callable function.

Adiabatic surface temperature

Heat flux gauge output can be converted to adiabatic surface temperature:

ghf = devices['"gauge-1"'].values          # kW/m2
ast = fds.astFromGhf(ghf, h=0.01, e=0.9, Tgauge=20.0)

h is the convective heat transfer coefficient in kW/m2-K, e the gauge emissivity, and Tgauge the gauge temperature in Celsius. The result is in Celsius, and equals Tgauge when the gauge flux is zero.