Almost all civil engineers use modelling for analysis, but are you using it right? This blogpost shares a few tips about it from personal experience.
Model or no model?
We must bear in mind that modelling is only a means to an end; it is only a way to arrive at certain conclusions so we know what to specify in our design. Whether done by modelling or not, and no matter how sophisticated the process is, it all translates into real materials and construction on site.
Before we proceed to establish a model, we must pause and think what results do we need in the end? How accurate do the results need to be? Check if there is sufficient data, and with sufficient accuracy, to back up the model. Modelling is an accurate way of predicting behaviour in a complex situation. But when the input has a large amount of uncertainty, accurate modelling is pointless. Rubbish in, rubbish out.
In many situations, closed form equations will be sufficient.
Even when we determine that we definitely need to resort to modelling, there are many different levels of sophistication or complexity in modelling.
To what degree of complexity should my model be?
A fundamental principle to bear in mind is that we, the engineers, must be the master in the modelling process. We drive the modelling process rather than being driving by the model. Simply going with whatever the computer tells you is one of the most dangerous mistakes in engineering designs.
It is easy to lose control of the model in a complex situation so when determining the degree of complexity of the model, the rule of thumb is to go with ‘as low as necessary’.
We must know what we want to achieve or get out of the model before even starting a model.
Always prefer 2D analysis over 3D
Reduce the model by using symmetry or zone of influence
Use linear model where required accuracy permits, but know when it is worth using non-linear models, including:
Geometrical non-linearity
Material non-linearity
Support non-linearity
The key to assessing when it is worth using some of the non-linearity is the benefit-cost ratio. Do it when the benefit significantly outweighs the cost of implementing it
What to do when undertaking complex modelling
Even when there are compelling reasons to undertake complex modelling, the process needs to be managed and developed with extreme care.
A rookie mistake a lot of modellers make is to try and get to the end in one go, hoping all will go well. This rarely happens and you often end up losing control.
If the model ends up with any bugs, it is nearly impossible to de-bug the model as all elements have tangled up and you can not tell what exactly caused the problem
In case the model can not be de-bugged, you are sometimes forced to start a new model from scratch. This will be extremely time consuming
A huge and complex model is nearly impossible to verify as there are too many factors at play
It is quite normal for an FE model to appear functioning perfectly well but give you a bunch of wrong results.
When undertaking complex modelling, follow the principles below:
Break the modelling process down to many different steps. Each step should be incremental on top of the previous step. Each step should be fully checked, verified and tested to confirm its correctness, before moving onto the next step. Therefore for each step, you should know exactly what results to expect. Use this expectation to fully test each step before moving to the next step
Use an umbrella structure of multiple models, with clear lines of communications linking all models. For example, one global model feeding into a number of complex local models. Only use large/global scale models where there are significant interaction between different parts of the model.
Comments