Hi all,
The cellspace contact detection logic in UDEC is invoked with “block config cellspace n1 n2” where n1 and n2 refer to the number of cells in the x- and y-directions.
However, some examples show that the the n1 and n2 parameters seem to be optional.
My question: how are the default values n1 and n2 calculated (i.e. what are the default values) if they are not manually specified in the command?
Cheers
Hi Zack,
The cell space default is 10x10 if nothing is specified. You can also see how many cells that you have in your model (if any) using the BLOCK CELL LIST command in the console.
If you run using the classic GIIC (UI), the initial set-up dialog shows this. I’ll make a note to update the documentation so this is clearer. Thanks for your question.
Cheers,
Dave
Thanks Dave.
I am running a model of a slope with the rough dimensions in the sketch attached. Horizontal width in the order of 100m. A region of interest close to the slope is discretised into bonded trigon blocks with edge length of 0.6m; the bonded block assembly is then overprinted with joints and bedding discontinuities.
Do you know, in broad terms, what will be the effect of using either (1) default cell spacing of 10x10; or (2) some finer spacing, e.g. 1x1 or similar.
Hoping to understand whether this will have any impact on the model in terms of run times or emergent behaviour etc.
Thanks!
Hi Zack,
Of course, the advantage of using a cell space is to much more quickly determine likely contacts for detection based on proximity rather than checking every possible contact. Of course this needs to be balanced as more cells means more overhead calculations. Note that the larger the cell space range (e.g., 100 x 100) the higher the cell space resolution. A cell space that is 1x1 is the same as not using a cell space.
Probably the best guide is based on what ITASCA’s PFC software does for cell space logic. PFC uses an automatic, adaptive cell space logic that is based on a study that showed that 5-20 piece extents (objects) per cell are optimal with little performance sensitivity within this range (learn more).
As a result of the simulations above, the target number of objects per cell [in PFC] is fixed at 10. Once the condition of more than 10 pieces [objects] whose centroids fall within a cell is detected, that cell is tested for refinement. Once fewer than 2 pieces map into a cell and that cell is a child cell, that cell is unrefined.
This being said, it would be good practice to play around with a few different cell sizes with a smaller, simpler model to see if you can fine tune your UDEC model. Also, from your diagram, it will ultimately depend on if that large, single block is a BBM (bonded block model) with the potential to break into smaller pieces or not. Note, you can remap the UDEC cell space once you’ve started the model.
And finely, note that while the cell space detection logic in UDEC does allow for the analysis of “flying” blocks, it is not efficient for problems with thousands of small particles. For something like that, PFC would be more effective.
Cheers,
Dave
