Debugging a Model.

The ill-condition error is the one we dread the most. One which can take days to solve, if we ever find it. Now I won't be able to help you with clearing this particular error at a snap. There are certain modelling techniques I follow to ensure an error free model.

While modelling the structure, connecting all the elements node to node is a must. This is an FEA software, you need the transfer of reactions whether it be loads, deflection or moments. As per my experience in RC structures, most of the errors/warnings can be solved through this. Minimize use of the option of all floors or use with caution. It helps when others work on your model too.

There goes the modelling basics, now let's get to the part of debugging. There are various kinds of errors, I might not remember all of it while I am writing this blog but I will try to answer the best I can.


Well for starters, standard solver would be the best. Although it takes much longer to analyse, it does give the error locations. You can set one load case to reduce the duration. 

At times, the standard solver is not of much help either. It gives non-existent coordinates, or one which really doesn't make any sense. My blog is particularly for this kind: 

  • Run the model. Check the type of error. One can get away just by knowing the different kinds of error, it requires experience though. 
  1. One example is, a mass source error usually associates with the diaphragm assignment. 
  2. The ~___ in the image below indicates a meshing error.

Notice the ~ in the labels.

  • Check the deflection animation. This will give you the anomaly in the model. Go for this first, rather than the standard solver. It will save time. An unconnected/orphan element will always displace differently, always.
  • An Eigen modal analysis is helpful too. You can check the contribution of each modes and figure out the anomaly. CSI recommends a dozen modes should be enough.
It might happen that the above doesn't help you either. I follow certain steps for it too.

  • Export the model as an e2k/f2k file for ETABS/SAFE. Import it back again. I don't really get why this helps but I guess it has something to do with the cache thing.
  • Delete portions of the model and run. Suppose you have a X story building. Run the base floors first, one at a time by deleting the other floors, till the top floor. This can get tricky and frustrating.
I once created a model with basement walls. And some portions of the wall (I had meshed it manually) had tension (FZ) reactions at the joints. A senior boss helped me figure it out. All the other sections were deleted and we worked only on that particular area. After many iterations, it was found to be a software issue. Re-modelling the walls and meshing it again was just fine.

So yeah, it does happen at times. And we have to figure it out as an Engineer.

Bonus point : Use Multi-threaded solver to check if your model has errors. It takes less time. Some errors can be found only after the analysis. For errors which show after creating an analysis model, just cancel the analysis and you can find the errors.


I will probably keep updating this blog as I learn more debugging techniques.

Comments

Post a Comment

Popular posts from this blog

Ritz and Eigen.

Image
Ritz, not Rizz and Eigen Vector are two ways of solving a Modal Analysis problem. It is what ETABS use while doing the Modal Analysis. Today's blog will be about the difference between these two. EIGEN-VECTOR : It determines the undamped free-vibration mode shapes and frequencies of the system. What free-vibration means in our term is that there is no external loads acting on it for any period of time, just the initial conditions. The idea of bending a ruler and allowing it to reverberate. The natural modes obtained through this provide an excellent insight into the behaviour of the structure.  But remember, it's the free-vibration response while the structures will be under various different loads. What should be done then? Here comes the Ritz Vector. RITZ-VECTOR : Research has indicated that the natural free-vibration mode shapes are not the best basis for a mode-superposition analysis of structures subjected to dynamic loading : Wilson, Yuan and Dickens, 1982. I stole this l...
Read more

Modifiers.

Image
This is one of the coolest concepts I have properly learnt at my job. I do know I will learn lot more on this. It's the Stiffness Modifiers.  It is a modification factor that is generally used in the properties of cross-section such as Moment of Inertia, Shear Area, Torsional Constant etc. It can also be used for Mass and Weight as per requirement. Modifiers for a typical Beam. Now why is it used? As the loads act on the structure, it will tend to generate cracks on the tension zone. Thus, reducing the area of cross section and stiffness of the member. 1 means the structure would take all of the loads, while 0 won't be taking any. It is kind of a paradox; the structure would resist all the loads if it is not cracked, and we know it will crack. In other way, we want the structure to not take the load as it will eventually crack. Concrete is weak in Tension. As per codes, IS Code ACI Code Now I have two great examples for putting my case. One is related to a torsion modifier I st...
Read more

Slab On Grade in ETABS.

Image
First let's know what an SOG is. The full form is Slab on Grade/Ground. One use for it to create an equal level for construction. Imagine a large area foundation with minimal area at a lower level. It is easier to provide reinforcement in a single plane; hence the lower level can be raised by filling it with soil or some other substances. Then it acts as one level foundation. I came across a scenario where I had to model an elliptical tank like structure, no columns in between. It was to be filled with soil. As you can see, I have attached an image, the slab span is quite large in between. The plan of the model. Imagine an ellipse with R1 = 37 m and R2 = 65 m. Now if you run the model as a normal slab, which it kinda is, one can easily guess the response. You can guess with experience that is, or a different case if you are a genius. I will tell you how the response would look like if you run it as a normal slab. You would get lots of local modes (I am assuming you know what local ...
Read more