This python module can be used for calculating the optimal sensor placement based on sampled FE solution data. It uses the numpy methods linalg.svd() and linalg.pinv() as well as the scipy method linalg.qr() for doing the main core of the computations.
Usage:
python3 sensor_placement.py <datafile> <num_POD> <num_sensor> [<nX>|<pixelfile> [<image_idx>]]
python3 osp_noplot.py <datafile> <num_POD> <num_sensor>
The first script sensor_placement.py works for 2D problems only,
where the values in <datafile> are samples in a Cartesian grid.
The parameter <nX> is then the number of data values in horizontal direction in each time frame,
and the number of values in vertical direction (nY) is derived from the total number
of values per time frame divided by <nX>. Alternatively, the <pixelfile> can be
specified which gives the index into a compressed array for each time frame,
or zero if the pixel is outside the rectangular domain. The <datafile> is
assumed to contain one time frame per line. This script will also
plot the results and create animations using the matplotlib module.
The second script osp_noplot.py can be used for any data set - also 3D, and contains no plotting. It only computes the optimal locations of the specified number of sensors, the reconstructed data field based on the sensor values, as well as the error (difference between the original and reconstructed data fields). All these calculations are also performed by the first script.
The results from the second script are saved to ASCII-files for visualization by external programs.
For instance, if the <datafile> here contains the time history of a nodal result quantity,
the result files will have identical topology and can be loaded as nodal result fields
into the program that created the <datafile> for further processing and visualization.