kowalskiw / fireeng-tools Goto Github PK

Edges mapping does not work. Currently only fully supported areas are supported, relaxing one of edges is not possible.

Need some efficient solution here.

Results of torsional analysis are not copied to the first TEM files of each profile.

Seems like missing call for Thermal2D.insert_tor() in the Mechanical.run().

Script does not work when there is no BEAM elements. The following error message is printed:
Traceback (most recent call last): File "D:\fireeng-tools\structures\iso2nf.py", line 661, in <module> run_user_mode(n, args) File "D:\fireeng-tools\structures\iso2nf.py", line 598, in run_user_mode m.make_thermals(arguments.config) File "D:\fireeng-tools\structures\iso2nf.py", line 419, in make_thermals tems, tshs, t_end = read_mech_input(self.input_file) File "D:\fireeng-tools\structures\iso2nf.py", line 74, in read_mech_input if len(tems[element[-1]]) < 3: KeyError: '1'

To be fixed and tested for structures with SHELL elements only.

Check points to be constrained in dummy structural analyses. It will allow to calculate reactions properly when area has some free edges.

One solution is to have exactly matching points at edges of meshes between structural files. The other is to implement some arc interpolation.

Elements at the boundaries of transfer domains are not calculated when the FDS cell size is higher than BEAM element size.

This is because we define transfer domains on cells centers. Thus BEAM elements that have size lower than the distance between domain boundary and center of the boundary cell are not assigned to any transfer domain.

While running SAFIR thermal analysis using CFD results, only one transfer file can be used. There is possibility to use more than one file. It requires setting different section properties for elements located in different transfer domains. To run thermal analyses it is required to rename transfer files to 'cfd.txt' one by one. This process seems to be perfect candidate for the new script.

There is what manycfds.py should do:

1. Modifying mechanical input file.
2. Copying and changing thermal input files to 'CFD' thermal attack.
3. Iterating over transfer files:

• finding the transfer domain boundaries;
• assigning modified in step (2) sections to elements that are located in the transfer domain boundaries;
• running thermal analyses of CFD-modified sections.

Wide testing is required to check possible and common errors.

The script should be used in iso2nf.py if more than one 'cfd*.txt' file is present.

Full documentation (in-the-file at least) should be developed.

1. Check the compatibility of the script with different FDS and SAFIR configurations.
2. Clear the code to be more object-oriented and more self-documented.

iso2nf.py repairs relaxations notation in xml files in thermal analysis, which is not needed. Only needed in mechanical analysis with relaxations.
This leads to copying sometimes multi-GB files for no reason.

There is a need to make iso2nf.py compatible with new version of SAFIR. Changes that affect the script are connected with new method of handling torsional analysis results while using natural fires heating. This new method has been introduced in v2022a0 of SAFIR.

The area2lineload.py script from structures is terribly slow. It took c.a. 30 min to run an example structure with 10k beam elements and 360 different load areas. And there were no edges routine implemented, only fully supported areas.

IMHO there are some bottlenecks, which I listed below. This improvements have to be done ASAP, to make the script practicable. Nevertheless, simple cases still can be covered with the current version.

1. Get meshing from .IN file - when the structure will be meshed exactly in the same pattern (same edge nodes) there will be no need to iterate over reaction nodes to find the two nearest points for interpolation. Resolution of the mesh would be exact - no overestimation and shorter SAFIR calculations. It can cause some bugs in gmsh API.
2. Less timestep for dummy shells - there are 10 timesteps, 1 s each to be calculated for a dummy shell. This value could be lower. However good reaction estimation should be achieved. 3 time steps seems like a safe and low-cost solution.
3. Assign summed loads - load from each dummy is assigned in single LOAD command. Thus, it is possible to visualize only the first load case (LOAD command). To see the sum of loads implemented by the script, it is required to sum it per element.

Any help with these would be very appreciated, as well as some other weaknesses of the code.

The script doesn't repair xml files after finishing the simulation with some error messages, as can be seen on the screens below. The simulations got to the end of the set simulation time.

Scripts have to be tested on Linux machines. We have already made some tests with iso2nf.py with Ubuntu 18.

There is a branch dedicated to improvements focused on Linux compatibility - unix_style.

Any help from Linux users would be more than appreciated.

How to introduce not planar surfaces with gmsh?