[anshul_UVA]

Hi All,

I just wanted to share an article that appeared in Autocar Professional's Aug 1 2010 edition. It shows some work that we do at India Science Lab (a lab associated with GM R&D). I am associated with the structures and safety group and was very glad to see such articles appearing in print about our organization and specifically the safety work (which I am associated with).

The article showcases some of the Human Body Modeling related work and where it can see potential applications in safety. This is a deeply involved technical field on which a consortium of Universities and OEMs are working. I would like to add that though the article may have made some relatively big claims on how the models can replace test dummies, this technology has a long way to go before we see it being widely accepted in test regulations.

I just wanted to share this with all the members here. This seems to be the kind of forum where members will have the right kind of questions and productive discussions can emerge from such a thread. So, if members have general questions about the safety work and/or other stuff they may want to know about R&D in automotive industry, they could ask me. I will try to answer them to the best of my knowledge (or take them to the more experienced members in my group).

And to add, this is not a publicity/marketing stunt. I was just very enthused seeing the article and wanted to show it to this community. So mods, if it violates rules, please remove it or if you think this is appropriate in another place, move it!

Hoping for a good discussion!

Regards,
Anshul

[Rehaan]

Hi Anshul_UVA,

I've had this thread open in my browser for more than a month. Finally got a chance to read the article, and am surprised noone else has commented as yet!

So it seems that computer simulations are the way things are done for a lot of processes in almost all industries these days, but yet the simulations aren't accurate or trusted enough to totally replace Anthropomorphic Test Devices (ATDs) when it comes to crash testing cars.
Ofcourse this is something you're working towards, but where is the major difficulty? From the article it sounded like the physics aren't an issue -- but modeling the "people" to behave realistically is. Whats the hardest part about this?

Also, all the ATDs i see are quite similar. Do manufacturers currently run tests for a vairety of body types (eg. obese person, short person, etc)?
I guess with the virtual testing this would be a lot easier to do.


I liked the part about IVHM (Integrated Vehicle Health Management) -- ie sensors that monitor values and predict a possible problem. Similar to the S.M.A.R.T system in computer harddrives.


Do take a look at Ajmat's thread on the GM Tech center visit as well :
http://www.team-bhp.com/forum/techni...bangalore.html

cya
R

[anshul_UVA]

Originally Posted by Rehaan

Hi Anshul_UVA,

So it seems that computer simulations are the way things are done for a lot of processes in almost all industries these days, but yet the simulations aren't accurate or trusted enough to totally replace Anthropomorphic Test Devices (ATDs) when it comes to crash testing cars.
Ofcourse this is something you're working towards, but where is the major difficulty? From the article it sounded like the physics aren't an issue -- but modeling the "people" to behave realistically is. Whats the hardest part about this?

Also, all the ATDs i see are quite similar. Do manufacturers currently run tests for a vairety of body types (eg. obese person, short person, etc)?
I guess with the virtual testing this would be a lot easier to do.

Hi Rehaan,

Well, firstly, thanks for the reply. I had given up on any replies and stopped checking this thread. So surprise today!

Well, About human body models, its really tough because even the basic physical aspects of human body are tough to model.

Body geometry - Even to create models, first we have to geometrically model all the body parts - even the small ligaments etc. This seems to be under control today mainly because of the amazing imaging technology available. MRIs give a whole lot of rich information of all that is inside our body.

Material models - Now that we have the geometry, the parts need to be given material models. Tissue behaves different when compared to the muscle which is different from the bone. Additionally, these properties change with age etc. So basically there are huge bands for material properties for the materials that the human body is composed of.

Interactions between parts - Even if geometry and material model is present for each part, a huge effort will go into defining interactions between parts of the body. For example, while the leg moves, our ligaments move and align around the joints to create the motion. All these (thousands of) interactions have to be properly modeled.

Validation - You can have a model but how will you validate it? eg. We would need a human test to validate a test for a 30 kmph impact on the head. For that we need a human willing to get himself impacted by a huge stone and let somebody study his response. Well, jokes apart, this is taken up mostly through cadaver testing, but there are huge legal restrictions on that.

These are some of the key kinds of issues research in this area, on which a lot of work is being done today. We might be close to getting a good human body model, but to get it into the main stream of testing and designing will take a huge mind-set shift.

About your second comment, about various sizes of ATDs, yes there are various sizes of ATDs. There are dummies representing a large male (95th %ile) and average male (50th %ile). Female dummies also come in the 5th and 50th %ile sizes. There are dummies for kids and infants too. You can see a nice pic from Wikipedia




But your comment that with modeling, the issue of size is easily handled is not entirely correct. Frankly, if you have created one model, you just cannot scale it up or down to get the model of a larger or smaller person. It is not impossible, but not very straightforward. various body parts need to be scaled in different proportion etc.

But these models are a fascinating prospect because the level of information these models can provide in crash testing can potentially reveal so much more about the injury sustained by the user than the dummy ever can.

There are a lot more thoughts floating in my mind right now, but I will restrict this reply to this. Do put in more comments so we can trigger some more minds.

Regards
Anshul