Build multi-component model with linear and non-linear tool
response functions
Save and easily re-apply models to new data
Easily transfer models to other Projects
Solutions
Sequential Quadratic Programming: this is a very powerful
family of robust, non-linear optimisation methods
Tarantola-Valette Direct Inversion: a Bayesian solution
technique taking account of a priori information to improve the
model and characterise uncertainties better in both logs and
parameters
Markov Chain Monte Carlo simulation: reduces the solution
space to a sub-region of maximum likelihood, then samples within
the sub-region to characterise the uncertainities
Model design
Models may be defined as single mineral sets per zone
Multiple mineral sets per zone can also be established with
sets switching automati-cally according to a partitioning curve that
changes as the log facies change
Data partitions are best created by SOM or can be user
specified; their role is to characterise the detailed geology more
completely and control mineral set switching
User control of tool uncertainties
Solutions constrainable against a priori information e.g.
XRD data; both single component and multiple component volume
constraints are possible
Output
Detailed automatic Layout Plots, can be fully customised by
the user
Unique Array-histograms clarify data relationships by
plotting all the components against inpout log data or log data
resuiduals
View residuals against inputs and results either as
histograms or with the unique array-histogram tool.
