| Contact Us | Site Map | |||
  |
|||
| Home | Software | Publications | Company | Customers |
|
Java Minimum Curvature Algorithm Java Minimum Curvature Gridding builds faulted, grid-based surface models with up to 30,000 rows and 30,000 columns from hundreds of thousands of data points, and hundreds of fault traces. It does this faster than any competing algorithms. The minimum curvature algorithm is hierarchical, starting with a grid with 2 rows and 2 columns. Though repeated refinements and biharmonic smoothing, it produces an output grid at the user specified row and column spacing. Even on computers with many gigabyts of memory and 64 bit operating systems, Java heap space may not be able to hold a grid with more than 2,000 rows and 2,000 columns in memory. Even if possible, holding larger grids in memory may results in memory "thrashing" which slows execution. To avoid this, JMCG can partition the final grid into up to 256 smaller sub-grids. If partitioning is required, after each refinement and smoothing iteration the sub-grids are mergered along overlapping edges. As a result of the repeated merging at each iteration, visual difference between a surface model done with and without partitioning is impossible to detect, and grid statistics indicate only small differences at the 7th significant digit. The example below was done on a PC runnning Windows 7 which has 3 giga-bytes of memory and and dual core AMD CPU. The image below shows the surface model generated from 500,000 random samples in four fault blocks. At the right is a video of the execution of the JMCG on this date.
| Real Time Example Running Java Minimum Curvature Gridding
Press Play to model 500,000 random points in a grid divided into 4 fault blocks using 3 fault traces with 9000 vertics. Running time is 18 seconds on a computer using a dual core AMD CPU The resulting surface model is shown at the left below
|
|
Copyright © 2011 Austin Subsurface Modeling, Inc. | All Rights Reserved. |