Preparing for use

Please go through the following steps to make SINAS suite running for you:

Preparing the working directory

To make sure that the SINAS IV PATRAN PCL menu appears within MSC.PATRAN, you have to copy the file p3epilog.pcl from SINASIV_BIN to your working directory. To do this follow the steps below in which <work-dir> represents the path to the working directory:

In addition to this, your working directory must also be used by MSC.PATRAN as the working directory. A recommended method to achieve this is to copy the desktop shortcut for starting MSC.PATRAN to your working directory. Change the "start in" field of this shortcut to refer to you working directory.

Notes about the provided user manual

The user manual for the software is included in the zip-file.

The user manual contains references to a GUI that can help starting the various modules. This GUI can unfortunately not be provided. Therefore all modules need to be started from the command line in a Command Prompt window.

The user manual is also referring to two modules named 'amesa1' and 'amesa2'. These modules are only relevant in the case thermo-elastic analysis is foreseen to be run with the finite element code ASKA. Since ASKA is not a frequently used software tool anymore, it has been decided not to build these modules for Windows.

A last module that the user manual is referring to is called 'otnse'. This is a stand-alone module that requires as one of the input files a MSC.PATRAN neutral file generated by ESABASE of a thermal radiatve geometric model. This module is not provided, because ESABASE is currently not being used a lot for thermal radiative analysis. The same functionality is provided by the PCL-software that can be accessed through the SINAS IV menu in MSC.PATRAN. Various thermal radiative geometric models can be converted to MSC.PATRAN through TASverter (see an other page at this web-site). This allows the user to create, verify and update the correspondence (overlap) between thermal lumped parameter nodes and structural finite elements.

PATRAN global tolerance and use of thermal geometry in PATRAN

With the use of TASverter, thermal radiative geometrical model files can be converted to PATRAN session files. Various formats are supported by TASverter. The session file contains PCL commands to create the thermal radiative geometry in a PATRAN database. This allows for building and verifying correspondence between thermal nodes and structural finite elements in one tool.

During the construction of the geometry within PATRAN, PATRAN uses the "global tolerance". This tolerance is used among others to decide whether or not geometrical entities have to be considered coincident or not. This tolerance is also used by the PATRAN PCL functions called by the overlap detection algorithm. In addition when manual manipulations are foreseen with the thermal radiative geometry within PATRAN (e.g. to implement the cutting operations) the global tolerance value is also relevant.

The current version of TASverter (r2007-03-19) creates PATRAN session files that do not include any statements or PCL function calls altering the default tolerance. It is foreseen that this feature will be implemented in future releases of TASverter. For this reason the user has to add manually the PCL function call to reduce the default tolerance. This function call has to be inserted after the function call that creates a new database (uil_file_new.go).

A good global tolerance value depends on the size of the smallest item in the model. A recommended global tolerance value that works fine in many cases is 1.0E-5. The PCL function call to set this value is

            pref_global_set_v3( TRUE, "1.E-5", "SAVE" )

Note again that this function call has to be inserted after the function call that opens the new database (uil_file_new.go).

Overlap detection method implemented in the PATRAN PCL-software

The overlap detection method uses a tolerance that can be specified through the user interface in the overlap detection form. Note that this tolerance is different from the PATRAN global tolerance. With this tolerance the method aims for building two sets of structural finite elements per thermal node:

In addition the set of unrelated elements is build with elements that are not corresponding to any of the thermal nodes. These elements may represent structural parts that are irrelevant for the thermal analyses, but are relevant for the structural analyses.

The building of these sets as it is implemented in the PATRAN PCL makes use of PATRAN functionality to project FE nodes on surfaces. The following basic steps are gone through for each thermal node:

Generation of thermal data file (.tdf) with ESATAN

In the user manual is mentioned that the SINAS thermal data file can be generated from ESATAN with the subroutine WRTTEM, which used to be part of the installation package of SINAS. This subroutine is not longer compatible with current versions of ESATAN, therefore an alternative procedure has been developed to generate the thermal data file from an ESATAN run. This procedure is explained in below.

As a first step a csv-file needs to be created with ESATAN run. To achieve this the following statement needs to be included in the $OUTPUT section of the ESATAN input file.

        CALL PRNCSV(' ', 'T', CURRENT, 'NODE', 'temperatures.csv')

Note that this call causes a file with name temperatures.csv to be generated. This file name is just an example file name and the user is free to provide any other file name.

In order to convert this file into a SINAS IV thermal data file, you can use the provided (new) module csv2tdf. The usage of this module is:

        csv2tdf <csvfile> <tdffile>

Please be aware of a limitation in ESATAN version 10 in which the number characters per line in the csv-file is limited to 10000.

Troubleshooting convergence SINAS main module

The two most common convergence problems occurring with the SINAS main module are discussed below with possible resolutions.

Overdetermined constraints

The thermal node temperatures computed with e.g. ESATAN is imposed to structural FEM through linear constraints. Each constraints specifies that the average temperature of the FE-nodes corresponding to the thermal node is equal to the thermal node temperature.

To summarise the constraint relations

        sum(i) a_ij T_ij^f = T_j^t
        sum(i) a_ij = 1

In matrix form this is:

        A T^f = T^t

With A being a rectangular matrix with the number of columns equal to the number of FE nodes and the number of rows equal to the number of thermal nodes involved.

A situation in which an overdetermined set of constraints can occur is when the thermal node mesh is as fine as the strucural FE mesh and in many cases combined with a circular structure. The latter causes to impose an extra constraint at the beginning of the chain introduced at the end of the chain.

A way to detect this is to compute the singular values (sort of eigenvalues for a rectangular matrix). If there are singular values of 0.0 (or close to that) then we have an overconstraint system.

Badly conditioned conduction matrix

SINAS uses an iterative solver based on the conjugate gradient method with pre-conditioning. This pre-conditionioning comes down to normalising the conduction matrix and filtering out all small matrix elements. SINAS uses the control file parameter threshold for the filtering algorithm: all values less than threshold are not included in the solution.

In some case, e.g. when the conduction matrix is not well conditioned the filtering method throws away too relevant information that causes the convergence process to be very slow or not to work at all. In those cases the solution is to lower the threshold value. Be aware that this will be at the cost of a higher memory usage.