Discussion Forum

Posted: 8 years ago 4 nov 2016, 16:16 GMT-4

Hi I see nothing wrong in the model, and the results are for me correct, for a linear analysis. But, what is far more questionable is the physical sense of your model, you are putting 100N or 10 kgf on a um sizes cantilever ! If you would validate this by test, your beam would break far before you manage to apply that weight on it. A Si bean an exceptionally sustain about 1GPa stress, you are going some 1000 times above that. Furthermore FEM, by default is linear and the part is not "breaking". That corresponds to the hypothesis sin(alpha)=alpha and cos(alpha)=1 but this is only valid for small angles, or small relative displacements. You have large displacements, therefore you should at least apply a geometrical non-linear model (turn it on by ticking the corresponding solver check box). Then you will get something closer to a true case of a large deforming beam. So, I suspect that you missed the loads units, probably you where thinking of a load of 10 or 100 [mN] One caveat with COMSOL is that it does not tell us the default grid units in the graphics, therefore it's very easy to misread the true size of the geometry. Adding the units seem to be a major programming difficulty, as I have asked for this "simple enhancement" for now close to 10 years, probably this has also to do that it is us users that are liable (and not COMSOL!) if we miss read a model and make a part (bridge, plane, house ...) that fails. Part of the FEM activity is the VV&C (verification, validation and calibration) of our models, this takes time to learn, but is essential if you want to continue to make a living from model delivery :) -- Good luck Ivar

Posted: 8 years ago 5 nov 2016, 00:55 GMT-4

[QUOTE] Hi I see nothing wrong in the model, and the results are for me correct, for a linear analysis. But, what is far more questionable is the physical sense of your model, you are putting 100N or 10 kgf on a um sizes cantilever ! If you would validate this by test, your beam would break far before you manage to apply that weight on it. A Si bean an exceptionally sustain about 1GPa stress, you are going some 1000 times above that. Furthermore FEM, by default is linear and the part is not "breaking". That corresponds to the hypothesis sin(alpha)=alpha and cos(alpha)=1 but this is only valid for small angles, or small relative displacements. You have large displacements, therefore you should at least apply a geometrical non-linear model (turn it on by ticking the corresponding solver check box). Then you will get something closer to a true case of a large deforming beam. So, I suspect that you missed the loads units, probably you where thinking of a load of 10 or 100 [mN] One caveat with COMSOL is that it does not tell us the default grid units in the graphics, therefore it's very easy to misread the true size of the geometry. Adding the units seem to be a major programming difficulty, as I have asked for this "simple enhancement" for now close to 10 years, probably this has also to do that it is us users that are liable (and not COMSOL!) if we miss read a model and make a part (bridge, plane, house ...) that fails. Part of the FEM activity is the VV&C (verification, validation and calibration) of our models, this takes time to learn, but is essential if you want to continue to make a living from model delivery :) -- Good luck Ivar [/QUOTE] =========================================================================== Thanking you for this kind of help. yes i observed that applied boundary load is not feasible. but major thing which i have been observed that when generating plot between arc length and displacement then selecting boundary 4, along length then its give graph opposite but if we select boundary 11 along length then its give properly. any specific reason for selecting 11, except 4. regrards, anand

Posted: 8 years ago 5 nov 2016, 05:16 GMT-4

Hi I believe that has to do with the definitions used in COMSOL: - arc length is a value "s" along an edge pointing in a given direction that comes from the topological decomposing of your domains, so the "direction" of the arc length depends on which one you pick, over a boundary the arcs normally form a closed loop (mostly, exception in case of bifurcations). - what you should do to make it more "Cartesian" is to replace the "arc length" by an expression and use X,Y,Z whichever fits your need then the directions will flip around as you expect. You can turn on the arc direction arrows under the "View" settings, its often useful to have several "views", with such arcs, and selection id numbers, or automatic scales ... This applies also to 2D boundaries (surfaces) as you have up and down directions (read through Pr. Enzo Tonti's book or articles about "The Mathematical Structure of Classical and Relativistic Physics: A General Classification Diagram (Modeling and Simulation in Science, Engineering and Technology)". It gives you a good insight in domains, boundaries, edges, points and their directions and classifications, be aware his definitions of "Globals" differs from COMSOL definition though. When you define cut surfaces, or cut boundaries COMSOL chooses either its internal arc-lengths/surface directions selection or you can choose the underlying Cartesian directions, then if you plot deformations be sure you select the correct u,v,w directions (COMSOL will not do it necessarily as you expected it) which gives easily strange results as you notice here (I just got caught it took me a couple of days extra as I trusted the image and was searching for a model error in my the setup, but no it was only a representation reference I got wrong ;). Another caveat with FEM, it gives always impressive images, but nothing tells us that these are "real" -- Good luck Ivar

Posted: 8 years ago 5 nov 2016, 08:19 GMT-4

thanking you, In this same model, i want to do strain analysis but choosing for Y-axis by given selection then it is confusion that which one is more appropriate for simple starin analysis. Suppose If I want to plot in this model is Graph between Force and Displacement then it is possible?? What are the procedure for both please share with us. regards, anand

Posted: 8 years ago 5 nov 2016, 16:13 GMT-4

Hello Well force and displacement OK but which one your model has a geometrical extent, when you say displacement is it in one particular direction, and of a particular point, or the average of a boundary (surface). The same for the force, FEM gives you pressures, you must integrate this pressure over a given surface to get out total forces, but which area is the one you want to plot for. In such cases I normally use Definitions Variables, to create global variables by integrate "density" values over a given test boundary or area. or take some average displacement of a surface. Then you can make a line plot with the Force in the main entry and your definition of your displacement as the "Expression" for the horizontal axis With COMSOL, do not forget that u,v,w are all fields, these are correctly written u(x,y,z,t) in all generality -- Good luck Ivar

Posted: 8 years ago 11 nov 2016, 01:34 GMT-5

Hello Mr.Ivar Kjelberg, Thank you for a nice disscussion with you, Now I want to model this same cantilever with Piezoresistive doping. i tried many time but facing difficulty to model it. its give error continue. i would like to elobrate a littlte bit with you. First in the same cantilever, at fixed constraint side where stress is maximum, a piezoresistive resistor size(120x20x2micron) inserted on the top at same lable. choosing for piezoresitive ,material is single crystal light doped material and then piezoresistitiy boundary current model adjusting value of Nd. now one terminal is ground and at another one giving 3[V] but still get error. Please tell us some step and if possible then model a demo for me. your valuable suggetion is very important for me. Thank you, regards anand