 |
The following pages illustrate a first step towards the development of a system that
would allow geological models to evolve backwards in time. The method of interactive
evolutionary computation (IEC) provides for the inclusion of geological knowledge and
expertise in a rigorous mathematical inversion scheme, by simply asking an expert user
to visually evaluate different geological models. All that is required is a code that
allows the user to forward model a process and view its result. An example of extensional
faulting demonstrates the potential of the technique.
Genetic algorithm
Genetic algorithms (GAs) are a search method suitable for the inversion of highly
non-linear functions. Starting with a set of random solutions, these algorithms
progressively modify the solution set by mimicking the evolutionary behavior of
biological systems (selection, cross-over and mutation), until an acceptable result
is achieved. Since GAs work by optimising an ensemble of solutions, unlike other
inversion algorithms that optimise one single solution, they are an obvious choice
as the internal engine for interactive inversion applications.
Method
Our IEC system works by linking a geological forward model to a GA. The forward
modelling code used here is a particle-in-cell finite element code which is well
suited to problems involving very large deformation.
Details
of this code can be found on the World Wide Web. The inversion
process works as follows: a geologist uses the computer code with the aim of
producing a geological model that matches a target geological section. A number
of selected parameters is allowed to vary within given ranges. The GA initially
generates a suite of different models using randomly picked parameter values. In our
case, these models could be static geological models or animations showing time
evolution. The geologist ranks each result according to geological criteria, guided by
his or her experience and knowledge. Once the results are ranked, the GA applies
mathematically rigorous methods to generate a new set of models that progressively
converges towards the target geological section. An element of randomness is also part
of the approach, allowing unexpected models to be generated and perhaps suggesting new
possibilities outside the experience or expectation of the geologist.
|