Solid modeling can sometimes be very frustrating for users – there are project dimensions to fiddle with, and a whole lot of settings to adjust. The process is particularly frustrating when you’ve waited patiently for a model to be displayed as an isosurface in RockPlot3D, only to see a blob that doesn’t remotely look like what you expected. This topic is designed to offer some general suggestions.
Missing Hot Spots
We get a fair amount of technical support emails concerning solid models that don’t represent hot spots very well – areas where a small number of sample values are a magnitude (or more) higher than nearby samples. Here are some general suggestions of how to approach solid modeling if you have values of (for example) 10,000 only a few feet away from values of 10 or 1,000.
- Logarithmic modeling will help to confine highs to a localized area.
- The modeling method you choose will make a big difference.
- Do not use Inverse-Distance Isotropic. This method uses all data values to interpolate a solid model node value, thus averaging the highs.
- Try Inverse-distance Anisotropic, which uses only the closest control point in each sector around the node to interpolate the node value. This helps reduce the averaging effect of low values on a localized high.
- Or, try Inverse Distance Weighting if it’s helpful to apply a different weighting exponent horizontally versus vertically. For example, if the low values are directly above or below the high, then biasing the model horizontally will minimize their effect on averaging the high value. Remember that to bias the model horizontally you enter a horizontal exponent that is LESS than the vertical exponent. To bias a model vertically, you enter a horizontal exponent that is greater than the vertical.
- The model dimensions, node spacing, and node locations will make a difference: how close are the model nodes versus the disparate data samples? If a solid model voxel contains two control points with disparate values, you will never get the high to be honored. Either tighten the node spacing or, if the model is too huge as it is, see if offsetting the nodes slightly (adjust x and/or y min) results in the high being alone in a voxel.
- The high-fidelity option should help, but (again) not if a high and low control point share a voxel. They will get averaged.
- How many instances of 10,000 are present? If you would expect only a couple of solid model node to be that high, you could use the solid model Range Filter option to bump the not-high-enough value up to the 10,000. First, run a Solid | Statistics | Histogram for the model to see the distribution of the interpolated values. If it looks like there is a very limited number of high values which are not quite high enough, you could boost them. Use Solid | Filter | Range Filter and activate the High Stop/Low Pass option to “stop? the high values (set the threshold to the not-high-enough values in the model), with a replacement of the desired (higher) value. If you do this, you might want to be sure to create a histogram of the source data values as well (I-Data / Histogram) should you be asked to confirm why you are messing with the model in this manner.
- There is an interactive 3D solid model editor which can be used to edit individual nodes or groups of nodes.
- There is also a 2D solid model editor which will pull a single horizontal layer from within a solid model and display it in an interactive editor where you can manually edit the node values. When you close the editor, the slice is re-inserted into the model.
Missing Low-Value Control
Another problematic scenario arises when grade samples are recorded for specific depth intervals, but below-threshold samples have been discarded and are not present in the database. If you know that non-listed depth intervals have low values (rather than not having been sampled at all - this is an important distinction!) then you can make use of the data resampling options to add control. Use the Add Dummy Values setting to add low-value control points.
RockWare home page