gfabieno/ModelGenerator

ModelGenerator

Generate pseudo-random physical ground models.

This repository is best used in conjunction with GeoFlow.

Installation

Clone this repository through:

pip install git+https://github.com/gfabieno/ModelGenerator.git

Contents

ModelGenerator pseudo-randomly generates ground models through the ModelGenerator class. Models are dictionaries of discrete, gridded representations of each of the defined properties. Models are generated by iteratively creating layers given a range of permissible geometrical features and given statiscal contraints for the properties.

The contents of this package are the following:

• ModelGenerator class: Generates pseudo-random layered model using generate_model, given selected lithologies. Holds all parent parameters.

• Stratigraphy class: A collection of lithostratigraphic sequences. Generates a specific sequence of layers through build_stratigraphy. (Layer objects hold the properties of a specific layer in a generated model.)

• Sequence class: An iterable sequence of lithologies.

• Lithology iterator: An iterable sequence of properties. (Property objects statistically describe a single physical property.)

• Diapir class: A specific lithology. Add a diapir-shaped deformation to layer boundaries.

• gridded_model function: Generates a gridded representation from a model depicted by a list of layers.

• Various functions and classes that implement specific generation mechanics of a model (Deformation, Faults, random_thicks, random_dips, generate_random_boundaries, random_fields).

Some of the implemented parameters are described in SIMON Jérome (2023) 3 Écarts de domaine en reconstruction de modèles de vitesse sismique par apprentissage profond: caractérisation de la transférabilité inter-domaines, Transférabilité et contrôle de qualité en estimation de modèles de vitesse sismique par apprentissage profond, Institut national de la recherche scientifique.

Sample usage

Examples are shown in ModelGenerator/examples.py.