Hi everyone,
Here is the problem description in the PDF inside the folder , and the Matlab code which can be used to show the error along a 2D plane.
In the other zip file you can also find the datasets
I have named them below, with the following conventions ...
Velocity05_it10 means, inlet velocity is 1.5m/sec and solution at iteration 10 (these are time instants). So,
Velocity13_it20 means, inlet velocity is 1.3m/sec and solution at iteration 20.
Using these datasets.. ), please use few of these datasets to build the Weighting Function (using PRBS method). Once the Weighting Function is complete, then, using only the inlet_velocity (the Weighting Function is a function of inlet_velocity), the nodal values are generated. The nodal functions I am looking at is U_vel, V_vel and W_vel (these are velocity magnitude along XYZ axes respectively).
To check what the error is,.. you can use the unused dataset to compare. To compare please use the generated realization and the acutal datasets. And, compare along the slices (use the Matlab code provided).
Please let me know if you have any questions. Thanks for the help. I truly appreciate.
Thanks.
## Deliverables
Here are some of the other details
In a typical airflow within a room, there exists linear relationship between the inlet velocity, the nodal values and their respective errors.
What I need to do is:
** If the inlet velocity is provided, the nodes (whose weighing functions are functions are inlet velocity) should generate the nodal U,V and W velocity. This velocity can be checked with actual simulation data.
** To use pseudo-random binary signals:
To ensure a robust model to predict near-zero (or reduces-to-zero) errors and that performs under real-time conditions, a simplified technique ??" the Pseudo Random Binary Signals (PRBS) is to be employed. In this project, the evolution of the inlet velocity should be used as a system Weighing Function for each node within the reduced volumetric zone. This Weighing Function will render the model robust in real-time, and reduce the error to zero by cross-correlation, or the error remains a constant (similar to a white noise), which then can be eliminated in real-time.
The text files can be used to build the Weighting Functions (to reduce errors). And, one file can be used as a TEST to check if the errors are less.
Typical dataset provided:
The datasets represent nodes in the following order:
Node number
X
Y
Z
Vel_Mag (velocity magnitude)
U (velocity along X axis)
V (vel along Y)
W (vel along Z)
T (temperature at node)
The files have been generated such a way that,
for example -- when the inlet velocity is 1.5m/sec from the window... it is represented as [login to view URL]; this means, the data what you see inside is taken during the 10th iteration (while waiting for convergence), and the inlet velocity is 1.5m/sec.
And then, you have to compare with the original dataset to quantify the errors. I have the Matlab function that "slices" the room into some 5 planes along X and 5 planes along Y axes (like cutting a bread loaf). You need to show the slices and the errors using the Matlab Function.
Thanks.