Simulates subsidence caused by longwall mining using a cellular automata model intended to by run on a supercomputer

MPI must be installed to run these programs - Installing MPI on OSX

Subsidence, the motion of a surface as it shifts downward, is often caused by mining operations that extract layers of minerals from underneath the earth’s surface. The displacement of the ground surface caused when a void created from mining collapses is both an environmental and infrastructure concern. A Cellular Automaton (CA) model for subsidence prediction iteratively simulates the collapse of a void using a grid of cells, each of which is a Finite State Machine that transitions between states based on the states of neighbouring cells. Accurate models utilize a large number of cells in a three-dimensional array - a computationally intensive simulation that is impractical to run on a normal machine. To mitigate the large computational times associated with such a simulation these programs explore the possibility of parallel processing using MPI and determine the extent that supercomputing facilities could improve the execution time of the model.

Animated 2D test program for visual analysis of CA behavior

Renders a Cellular Automata simulation in real time – one iteration of the cell grid per frame. Each cell type is represented by a different colored square: dark blue (coal), light blue (earth), black (drill), and white (air/void). The elapsed time is displayed in the top-right corner, although this is not an accurate representation of the required computational time (rendering slows execution substantially).

./anim2d –w 600 –h 110 –rx 2 –ry 2 –csh 5 –dl 240 –gh 100

Animated 3D test program for visual analysis of CA behavior

Renders a Cellular Automata simulation in real time – one iteration of the cell grid per frame. The Cell Grid is represented by a height-map, where the highest ‘Earth’ cell determines the y-axis location of each point. Vertex coloring also indicates height with lower vertices colored red, fading to green for higher points.

./anim3d

Parameters are specified as XML markup in Config.xml. All tags should exist inside of root tags. In regards to types, parameters expecting a character accept strings, although only the first character is used (allowing cell types to be named, but represented by a single byte e.g. ‘earth’ is represented by the ASCII value for ‘e’). Parameters expecting a Vector2 or Vector3 require two or three decimal numbers respectively, separated by spaces.

Displays results of MPI test run visually (2D)

This program renders the output of a CA simulation after its completion. Rendering is the same as in Animated2D. Computation of the CA uses MPI, although it is intended for Mac/Unix (not a supercomputer). This assumes that an MPI library is installed on the local machine. Distribution of computation to multiple processors is achieved using horizontal subdivision of the cell grid. The program must be run with at least two processors as one is used as a shared memory server. The final cell data contained on each processor is collated after the required number of iterations is completed – all processors send the data to the processor with an index of 0, which then displays the complete data set. This program is used to visually check that an MPI algorithm is working as intended.

mpirun –np 4 ./vismpi –i 1000 –rx 2 –ry 2

Generates a model file of a 3D CA simulation

This program outputs a high-resolution model of the ground topology that results from a 3D CA simulation. The model is in the format of a Wavefront OBJ file, allowing it to be opened and viewed using virtually any 3D modeling software. The program does not use MPI; it uses only a single processing thread.

./highres

A variation of Animated2D that allows for different material types.

Renders a Cellular Automata simulation in real time – one iteration of the cell grid per frame. Loads the initial configuration from an xml file that specifies cell types properties and an image file that defines initial cell states. Each cell type may have a different mean and variance for the probability distribution and kill-bubble probability. The image supplied is converted to cell grid values based on the pixel color.

./ca ConfigFile.xml

<CellularAutomata>
  <CellMap>CellMap2.png</CellMap>
  
  <Coal>0x03417A</Coal>
  <Drill>0x000000</Drill>
  <Air>0xFFFFFF</Air>
  
  <CellType>
    <Colour>0x4697DF</Colour>
    <ProbabiltyDistribution>
      <Mean>0.0</Mean>
      <Variance>3.0</Variance>
     </ProbabiltyDistribution>
     <KillBubble>0.002</KillBubble>
  </CellType>
  
  <CellType>
    <Colour>0x0A5394</Colour>
    <ProbabiltyDistribution>
      <Mean>0.0</Mean>
      <Variance>0.0001</Variance>
     </ProbabiltyDistribution>
    <KillBubble>0.0</KillBubble>
  </CellType>
  
  <CellType>
    <Colour>0x46b9df</Colour>
    <ProbabiltyDistribution>
      <Mean>-2.0</Mean>
      <Variance>3.0</Variance>
    </ProbabiltyDistribution>
    <KillBubble>0.0</KillBubble>
  </CellType>
</CellularAutomata>