Finding the optimum structure type

<< Click to Display Table of Contents >>

Navigation:  Tutorials > Modelling your own sample >

Finding the optimum structure type

All the files necessary for this tutorial are contained in the zip file Intermediate tutorial modelling own sample. (You may have already downloaded some of the datafiles, but they are bunched together again for your convenience.) Download the zip file from the PoreXpert website at  .Our results are included for comparison to yours, but may differ because they were carried out with learning mode turned on (the default setting) and our program is very experienced in modelling a wide range of samples.



Start PoreXpert. Then Sample (i.e. load) the datafile high_porosity_mineral.csv .  Click Accept.


You will now be at the Curve Fitting screen.  Instead of clicking Accept, to run a batch click Go home.


Click PoreBatch | Open batch description, and Open the batch file fitting-all-types-15x-batch.pXt .


Click Porebatch | Run batch list , and wait for the batch to complete.  (If you are running a computer with enough processors and threads, many or all of the processes should run in parallel.)


On completion, click File | Save as... | PoreXpert file , and save your results in case you need to look back at them.


You will notice that the vertically banded structure type is the one with the closest fit (smallest percentage Distance between simulation and experiment). So that is the structure type to choose. (If your own results differ from that, for the purposes of this tutorial choose the vertically banded structure type anyway.)


The next step is to generate at least five stochastic realisations of this structure type with different random number seeds to find the most statistically representative structure.  


Click File | Open ... , and reload the datafile high_porosity_mineral.csv .  Click Accept. At the Curve Fitting screen, to run a batch click Go home (not Accept).


Click PoreBatch | Open batch description, and Open the batch file fitting_batch_5_stoch_gens_VB_x25.pXt . Then PoreBatch | Run batch list . After completion, save the output as .pXt file under a different name.


You then need to choose the most representative structure, as explained in the advanced tutorial at .  However, for purposes of this intermediate tutorial, we will choose the stochastic generation which has the lowest Distance - in our case 2.09% for stochastic generation 1 .  The lowest distance can be selected using the Build operation in the next section of this tutorial.  However, you may wish to select the closest fit yourself by highlighting each operation with a higher distance %, and deleting it.


You are now ready to carry out a cluster analysis on your sample.