XANSYS Message: 38590 [Go back to message list] [bookmark on del.icio.us]
No rating yet Subject: Re: Optimizing section capacity Author: Yaser Jaradat Date: 2002-06-07 10:08:00Thank you shen_yeh for your concern. appreciate that for you.
What I am going to do is use P as a DV and any variable related to the P as an OBJ. This way it will work but the problem appears when I have to use P and thetas of the fiber angle at each layer as DVs. If I am going to use the stress criteria as my SV, then it is not going to work perfectly because you may reach a maximum stress at lower P because of the layer orientation. For example, if you have P applied on 90 deg. with fiber orientation, you will get a max. stress with a very low P. I think this kind of optimization configuration will not work...I have to think of other way to manage my variables.
Yaser Jaradat Dept. of Civil Engineering University of Maryland College Park, MD 20742 Phone: 301-405-1952 shenyeh_chen wrote:
> You can do this :
> Find DV : P0, X1, X2, X3.......XN > To Maximize : P > Subjected to : Damage_Parameter < Damage_Criteria > Stress < Allowable_Stress > Displacement < Allowable_Disp
> If you defind P0 by
> OPVAR,P0,DV,.......
> ANSYS will allow you to assign a initial on P0, yet it recognizes that > you have to assign an initial value , but (see example in VM157) > ANSYS will change it during optimizaiton iteration. I don't > know how can ANSYS recognize this, but seems like it DOES (probably > someone can tell us how ??)
> Then you control your load by assigning > P=P0
> Is it OK to assign the same value (althoug different process) into > both DV and OBJ (theoretically)? Definitely yes. The famous Feasible > Direction Method uses this approach. So it is just a question of how > to trick ANSYS. I use my own program so I don't have this problem. > You may have to do a little test.
> However, you can not have multi-objective. What you can do is you > maximize one objective, and constrain the rest of them. Or, you use > weighted sum on all objective and come up with one number. There are > some multi-objective algorithm available, but ususally they > also give multiple optimum, and unless you know how to weight each > objective, you still don't know how to pick the best one.
> --- In xansys@y..., Yaser Jaradat wrote: > > In my case capacity is the maximum load at which the stresses in all > elements are below the failure limit > > (within the failure criteria). It will be one case at a time, one > time max. capacity until damage happens, other > > time max. buckling load...etc. The load"P=10" is just a starting > value, I can put any value because the load is > > unknow, it has to be calculated. > > So the case is to get the highest load that the section can carry > such that stresses do not exceed allowbles.
> > Yaser Jaradat > > Dept. of Civil Engineering > > University of Maryland > > College Park, MD 20742 > > Phone: 301-405-1952
> > shenyeh_chen wrote:
> > > So, You want to optimize (probably maximize) the capacity of the > > > structure under the loading "P=10".
> > > Again, depending on what you mean by "capacity". It can be > stiffness, > > > can be vibration frequency, can be the maximum load it can take > > > until damage happen, or, as one gentleman posted previously, it > > > can be your buckling load level. You perform the analysis, get > > > this capacity number you want from the analysis, put into > > > a parameter, and defined this parameter as the objective function.
> > > You may say it is "everything" your want to maximize. But that is > > > not possible. Because you can image that, there is > > > possibly no solution that will make stiffness, damage, etc all > > > the maximum. If your objective function is more than one value, > > > either there is a "group" of designs that is better than all other > > > designs, or you add up all the parameter to define an auxillary > > > function.
> > > For example, if you want to maximize both stiffness and damage > level. > > > Let's say you have a design of the two parameters (10,15) > > > How do you compare to another design of (12,13) ? How do you > > > define which one is better mathematically?
> > > --- In xansys@y..., Yaser Jaradat wrote: > > > > I understand that logic. but this is the problem, I have a > section > > > of composite layers and I want optimize it > > > > for maximum capacity through fiber oriention....So what should > be my > > > objective function if it is not P ?
> > > > Also I thought of an other OBJ function other than P and related > to > > > P, like deformation or stress! > > > > Just I wanted to know if it works with P as an objective > function, > > > but it seems not.
> > > > Yaser Jaradat > > > > Dept. of Civil Engineering > > > > University of Maryland > > > > College Park, MD 20742 > > > > Phone: 301-405-1952
> > > > shenyeh_chen wrote:
> > > > > I don't think it will work.......
> > > > > P is an objective function, which means it has to be evaluated > in > > > the > > > > > process. In your input, the only action to evaluate > > > > > P is
> > > > > "P=10"
> > > > > So P will be 10 forever. A Constant. How can you > > > > > minimize/maximize a constant ?
> > > > > You will have to understand your problem more. The capacity > should > > > > > be a function of the load you apply, among other things (like > > > > > fiber orientation, ply thichness, etc) not the load itself. > > > > > Defining the load you are alreafy applying as a value to > optimize > > > is > > > > > not making any sense. You probably want to optimize some > response > > > > > caused by the load (???).
> > > > > > Don't worry about me, I will not get heart burned, I have a > lot > > > of > > > > > water to cool it ....:) ...Its just a matter > > > > > > of misscommunication.
> > > > > > Regarding my problem, when I said didn't work I meant that > when > > > I > > > > > assigned the applied nodal load as an > > > > > > objective function it didn't work ...what I wanted to know > is > > > if > > > > > ansys can accept an input parameter as an > > > > > > objective function for optimization or not. Or it has to be > a > > > > > variable to be calculated within the process like, > > > > > > Volume, Area, Stress, buckling load....etc.
> > > > > > Yaser Jaradat > > > > > > Dept. of Civil Engineering > > > > > > University of Maryland > > > > > > College Park, MD 20742 > > > > > > Phone: 301-405-1952
> > > > > > shenyeh_chen wrote:
> > > > > > > Yaser,
> > > > > > > Don't get heart burn because of this. If you are posting > on > > > > > > > fea-optimization or LS-DYNA group, you will be asked to > > > identity > > > > > > > yourself too (or you will be totally ignored). I even got > > > kicked > > > > > out > > > > > > > of this group few weeks back because my eamil address > looks > > > like a > > > > > > > free one. So I had to change my email to my company > address.
> > > > > > > Capacity of composite can be tricky, and you have to > define > > > what > > > > > > > you mean by "capacity" first. It can be the compliance, > can be > > > > > > > the stiffness, or it can be the damage level.
> > > > > > > There are a lot of situation that we can catagorized as > "not > > > > > > > working" in optimizaiton. Either it doesn't reach a > feasible > > > > > solution, > > > > > > > or it doesn't give an optimum solution, or it gives a > wrong > > > > > solution, > > > > > > > or it escapes from your desired design space, or it never > > > > > converges > > > > > > > and iterates forever. Which one is your situation ?
> > > > > > > remember that optimization is not only a nonlinear > problem, > > > it is > > > > > > > an "inequality" problem too. So the way you formulate it > will > > > be > > > > > > > very important.
> > > > > > > --- In xansys@y..., "Bohlen, Dan (GEAE)" > > > wrote: > > > > > > > > Sorry, nationality does count. Ansys is not available > in > > > > > certain > > > > > > > countries > > > > > > > > therefore people who post from those countries cannot > have a > > > > > > > legitimate copy of > > > > > > > > the program. We don't like to help those with cracked > > > copies of > > > > > > > Ansys.
> > > > > > > > On a less legal angle, it's just polite to know who you > are > > > > > talking > > > > > > > to.
> > > > > > > > You should have asked me this question long time ago > because > > > I > > > > > am > > > > > > > writing to the > > > > > > > > group almost everyday and I am > > > > > > > > getting responses most of time. Regarding who am I and > where > > > > > from, I > > > > > > > think all > > > > > > > > of us are using this group to > > > > > > > > share knowledge and information doen't matter the race > or > > > > > > > nationality. The > > > > > > > > response should not based on such > > > > > > > > thing. Also I am attaching a signature with my > information > > > in > > > > > all > > > > > > > my emails.
> > > > > > > > > If you tell us who you are, and where your are from, > you > > > will > > > > > > > probably > > > > > > > > > get more response.
> > > > > > > > > Optimizing the composite fiber orientation is an old > topic > > > in > > > > > > > > > optimization. You should do a paper search.
> > > > > > > > > If you are doing research on this field, you should > > > probably > > > > > try > > > > > > > to > > > > > > > > > figure out the formulation by yourself.
> > > > > > > > > > I asked this quetion yesterday about maximizing the > > > load > > > > > > > capacity > > > > > > > > > for a built up section by using the load > > > > > > > > > > which is an input parameter as and OBJ > variable.....I > > > tired > > > > > it > > > > > > > but > > > > > > > > > didn' work !
> > > > > > > > > > Any comments on that ?
> > > > > > > > > > Yaser
> > > > > > > > > > Nanjudareddy wrote:
> > > > > > > > > > > Hi
> > > > > > > > > > > Please cross check your material properties and > and > > > units. > > > > > > > > > > > we have run the model with different material > > > properties > > > > > and > > > > > > > found > > > > > > > > > the > > > > > > > > > > > conduction through out the length.
> > > > > > > > > > > >Thus said I'm sorry for not having followed the > > > nettique > > > > > > > about > > > > > > > > > > > >signing the mail. At least I said what the topic > was > > > on > > > > > the > > > > > > > > > subject. > > > > > > > > > > > >Now I repost my question. > > > > > > > > > > > >Thank you for your help.