What’s happening here regarding block sizes and variance
Variance decreases when block size increases.
How is a grade of a block obtained during DBSim
The grade of block is the average grade of all the points inside the block.
How much the variance of the small block v is different from the variance of big block V?
…
Name four (4) issues that DBSim overcomes? (Disadvantages of the traditional techniques)
- Size of the Domain
- Support problems (could lead to computational issues).
DBSim has to ability to handle data with different support - Efficient but SLOW
- Store all the data (takes to much space), when sequentially simulate point by point
Name six (6) general characteristics (fonctionnement général) of DBSim
*important
- It is a further development of GSGS, where computational efficiency is improved trhough simulation of goups of nodes sharing the same neighborhood . (instead of point at a time)
- Sequential
- Simulation is carried out simultaneously for a group of nodes, which coincide with the internal poitns of each block
- Simulated nodes are averges out into block value, and internal points are discarded
- Simulated value are obtained by solving a joint simulation, identical to Standart joint LU-simulation method
- Neighborhood is used for conditioning the simulation (existing datat & simulation blocks)
Effect of block size over simulation:
As the block size gets smaller, DBSim→???
As the block size gets smaller, DBSim→SGS
Effect of block size over simulation:
As the block size gets bigger, GSGS→???
As the block size gets bigger, GSGS→LU, but DBSim just looses variability
Name the eight (8) steps of DBSim
- Perform normla score of the data.
- Define a random path that randomly visit N blocks.
- Determine the coordinates of the K nodes inside the visited block and find neighboring conditioning data.
- Derive all the required covariances and cross-variances.
- At the given block, generate the simulated values of the K nodes discretizing the block.
- Determine the simulated block average in the original data space (store in a file)
- Average the K simulated values in the Gaussian space for the block to be simulated. Discard K nodes.
- Repeat 3-7 until all blocks are visited.
Name four (4) Advantages of the DBSim algorithm
- Major increase in memory allocation due the discarding of the internal points
- Subtaintial improvement in computational efficiency
- Major efficiency improvement in variogram validation at a block support scale compared to that.
- No distribution assumptions are made for the chage of support unlike any other method
Takeaways from this graph?
That performance is based on group size and conditioning data. (There is an optimum group size)
Case Study: Walker Lake
What? Why? Process?
What?: DBSim was only used to change the support size to blocks rather than points
Why blocks rather than point support?: The goal is to determine the conditional probaility distribution of ore blocks that are classified one by one
Process?: 78,000 points on a 1m x 1m grid→780 blocks of about 100-point grades
Case Study: Walker Lake
Explain what happened & why
Higher frequency of extreme values because less pairs are available to acalculate the experimental variogram than in point support
Case study: Walker Lake
Takeaways from this graph?
Study Case: Walker Lake
Takeaways from this graph?
At high grades, DBSim represents batter the connectivity of the data set than SGSim does.
What does Connectivity refers?
Connectivity refers to the degree to which different parts of a mineral deposit are physically linked or related to each other.
Describe this illustred step of DBSim
Initial Situation, where drillholes are represented as red dots and the nodes to simulate in yellow
Describe this illustred step of DBSim
Blocks of chosen size is determined, using a random path, and a search neighborhood depicted in red. Internal are simulated by LU to find the value of the block
Describe this illustred step of DBSim
The nodes of the precedent block are discarded, and we go to the next block liberating memory from storing all the nodes
Describe this illustred step of DBSim
New block, same process.
Covariance can be computed separately to to still save computational costs
Recall LU LU
Hi LU LU <3
