DomainMCF Help

Maptek DomainMCF is a simple and powerful resource prediction tool. It combines cloud computing and machine learning to generate resource models from sample data in substantially less time than traditional resource modelling methods.

Read on for an overview of the DomainMCF process. Otherwise, skip straight into the details of using DomainMCF with one of the following topics:

  • See Job Hub for an overview of the DomainMCF interface and all the tasks you can perform.
  • See Configuring and Running a Job for detailed information on how to configure and run a job.

Process Overview

The DomainMCF process is straightforward. Initially, you’ll need to supply your drillhole sample data as a set of points with associated domain codes. DomainMCF uploads the data to the Maptek Compute Framework (MCF) The Maptek Compute Framework (MCF) is a cloud-based service that performs the computationally intensive parts of the DomainMCF process. This includes the tasks of training a machine learning model on a potentially large input dataset, and predicting domain boundaries based on this input., which trains a model that can either be used immediately to generate a block model representing predicted domain boundaries, or downloaded for future use.

A single processing run on a set of input data is represented in DomainMCF by a job. Based on the input and output types you supply, there are three basic kinds of job, as follows:

  • Training only
    If you create a job and supply sample data only, DomainMCF will train a model that you can download as an encrypted trained DomainMCF machine. Because the training step is not deterministic, having a persistent representation of the trained model allows you to reuse the same model in multiple prediction jobs, yielding consistent and repeatable results. The persistent file is available for audit purposes and the job output is reproducible after job deletion.

  • Predicting only
    If you supply a previously trained DomainMCF machine along with parameters defining a block model, DomainMCF will skip the training and use the supplied machine to perform domain prediction and generate a block model representing domain boundaries. This is helpful when you want to experiment with multiple blocking schemas on the same sample data, because the training step will not have to be repeated in each job. You can download the block model as either a BMF or CSV file.

  • Training and predicting
    Training and predicting can be performed in a single job by supplying both sample data and block model parameters. DomainMCF will train on the supplied data and perform domain prediction. In this case, you can choose to download any or all of the outputs, including the trained DomainMCF machine or the block model representing the predicted domain boundaries, either as a BMF or CSV file.

After downloading the results of a predicting job, you can view and perform further analysis on the block model in other software, such as Maptek Vulcan or Maptek Vulcan GeologyCore.