NOTE:  Understanding the contents of this Mail will help you not only 
improve Seat Belt Performance in Dyna but will probably increase 
Computation Speed significantly in certain cases.

Those who do not use Seat Belt models in Dyna, may skip this mail entirely.


In late October 2008, I had sent out a "Learning Aid" that contained the 
Belted LSTC Dummy Model within a very simplified Sled Environment. The 
purpose was to demonstrate to users how to model this kind of Sled 
Environment and Dyna Belt.

These files were extensively annotated, to assist everyone in 
understanding the various principles behind it.

Those who did not get the original, may use the following Link to 
download it for reference:

http://ftp.lstc.com/user/lstc-dummies/LSTC.LEARNING_AID_VEHICLE_AND_50TH_BELTED_DRIVER.081029_V1.0.zip


However, I realized last year that if some small but important changes 
were made to the Belt Model, their performance could be improved quite a 
bit, as well as increase the overall "speed" of the computation.

Based on that, I have attached here a new Learning Aid Package, 
containing the Modified Belt Model. The new job submission deck "dyna.k" 
has the Belt Model, Vehicle Model and Dummy Model "included" in it.

If you "unzip" the package in a separate directory, you will find the 
following files:

DriverBeltModel.TemplateWithLoadLimitingPyroRetractor.Mod.k
VehicleEnvironment.TemplateWithPitchDropYawPulse.k
LSTC.50thRigidFE.PostFeb28.WithBeltVehicleModelTemplate.k
dyna.k

Please run "dyna.k" as the main input file during job submission.


The main changes to the Belt Model compared to the "original" are simply 
these:

1) I have increased "lmin" to 3.0-mm.    (Note:   It used to be "1.0-mm" 
before.)
Note that "lmin" is found in  *MAT_SEATBELT .
There are three occurrences of "lmin" in this model and I have changed 
all of them.

Please read the Manual for the definition of "lmin" in *MAT_SEATBELT.

Also, please read *ELEMENT_SEATBELT_RETRACTOR and try to understand how 
"lmin" affects the "Belt Payout", working together with "lfed" within 
the same card.

We suggest the use of   lfed = 10.0 mm   in most cases. That is what I 
have used in this model.

All Notes in the Manual for the above are important. Please read and 
understand as much as possible.

2) I have made all the Belt Segments (the single line belts) of equal 
length within a single stretch (i.e, Retractor to D-Ring, D-Ring to 
Shoulder, Lower-Shoulder to Buckle, Buckle to Lap and Outer Lap to Anchor).

I don't mean that all segments are of equal length but simply that 
"within the same stretch" they are equal.

To understand this, I have attached two pictures of the Belt Segment 
between Retractor to D-Ring, one zoomed in near the D-Ring and the other 
at the Retractor area. I have put both the original and the new belt 
pictures, separating them in space so that you know what I have done.

The one on the right is the new belt and the one on the left is the 
original.

The pictures did not come out all that well but you can understand what 
I mean by looking at the "yellow spots" signifying the "nodes" of the 
segments. The original had much smaller segments in those areas, whereas 
in the new one they are all larger and equally spaced throughout.


In addition to that, the "dyna.k" file has the following extra cards to 
"mass scale" and maintain a "constant Time-Step" :

$
$
*CONTROL_TIMESTEP
$            The following gives Natural Time-Step
$    0.000  0.900000         0     0.000  0.000000         0
$#  dtinit    tssfac      isdo    tslimt     dt2ms      lctm     
erode     ms1st
$    0.000  0.900000         0     0.000  0.000000   1000057
     0.000  0.900000         0     0.000 -0.003112   1000057
$
*DEFINE_CURVE
$#    lcid      sidr       sfa       sfo      offa      offo    dattyp
   1000057         0  1.000000  1.000000
$#                a1                  o1
$
               0.000           0.0028000
        2000.0000000           0.0028000
$


Because the above lines are situated after all the other files are 
"included", the *Control_Timestep here will override the occurrence of 
the same in any other file.

What I have done is to use "Mass Scaling" in this Model to increase 
Time-Step of computation.

Dyna will first "Add Mass" (if required) and raise the TS to 
0.003112-millisecond and then multiply it by 0.9 to make it 
0.0028-millisecond (or 2.8-microsecond).

However, the Curve 1000057 will also "limit" the TS to 2.8-microsecond 
(it will not let it go beyond that).

This way, a "constant TS" of 2.8-microsecond will be maintained 
throughout the run.


I have determined that use of Mass Scaling in Seat Belt Models in the 
above way does not degrade performance as I had originally thought. In 
fact, I now feel that it could actually improve performance and speed 
quite a bit.


Hopefully this will help all users to improve the performance of Dyna 
Seat Belt Models.

Please let me know if you disagree with anything that I have written above.


Thanks,

Sarba Guha
LSTC, Troy Office, Michigan, USA


Additional NOTES:

a) Please monitor Added Mass throughout the run by reading in "glstat" 
in LSPP and ensure it is never excessive. This is a General Rule for all 
runs involving "Added Mass".

b) If the Seat Belt and Dummy is used in a Vehicle Model that has a much 
lower TS (like 1.0 microsecond), you cannot use the above method of Mass 
Scaling to increase it to 2.8-microseconds. That is because it will 
surely add excessive mass to the whole vehicle model as well. You can 
still use the above cards but "limit" the TS to whatever the natural TS 
is for the Vehicle Model.