3D Induced Polarisation Inversion
To determine the capacity and maximum efficiency of 3D inversion commonly applied to mineral exploration, specifically related to DC resistivity and Induced Polarisation data.
Area of science
Systems used
Magnus
Applications used
DCoctreeinv, IPoctreeinvThe Challenge
The volume of geophysical data acquired in the mineral exploration process has increased significantly during the continual improvement of acquisition method and systems. To effectively make full use of these datasets in an efficient manner presents significant computational challenges. The size of raw data, need for dense mesh definitions and accurate topographic definitions create a high demand on RAM available. The necessity to complete the computations necessary in a reasonable amount of time (on the order of weeks) requires a large number of fast computing nodes.
The Solution
The CPU dedicated to a single survey will be adjusted to find the most efficient number of CPU’s that can be used and determine the scalability of 3D inversion algorithms commonly used in geophysics. The efficiency of this upper limit of CPU’s will be tested against varied input parameters (namely mesh size) to determine the possible accuracy that can be achieved.
The Outcome
The scalability of the DC Resistivity/Induced Polarisation code used was disappointing. It was found that the code did not scale well past 20 CPU’s which does not make use of the available computing capacity. It was found the code needs to be re-designed – essentially from scratch. However, available RAM did provide the capability to significantly increase mesh sizes and inversion area which dramatically improved the accuracy of results. This was found to be particularly important in areas of large topographic relief.
