Well Log Measurement Prediction Using Neural Networks with Keras | by Andy McDonald | Oct, 2023

An instance of predicting bulk density (RHOB) with Keras and illustrating impacts of normalisation on prediction outcomes

Giant quantities of knowledge are acquired each day from wells all over the world. Nonetheless, the standard of that knowledge can differ considerably from lacking knowledge to knowledge impacted by sensor failure and borehole situations. This will have knock-on penalties on different components of a subsurface challenge, comparable to delays and inaccurate assumptions and conclusions.

As lacking knowledge is likely one of the commonest points we face with effectively log knowledge high quality, quite a few strategies and strategies have been developed to estimate values and fill within the gaps. This contains the appliance of machine studying know-how — which has elevated in reputation over the previous few many years with libraries comparable to TensorFlow and PyTorch.

On this tutorial, we might be utilizing Keras, which is a high-level neural networks API that runs on high of TensorFlow. We are going to use it for instance the method of constructing a machine-learning mannequin to permit predictions of bulk density (RHOB). It is a generally acquired logging measurement, nevertheless, it may be considerably impacted by unhealthy gap situations or, in some circumstances, instruments can fail, leading to no measurements over key intervals.

We are going to begin with a quite simple mannequin, that doesn’t account for normalising the inputs, a standard step within the machine studying workflow. Then, we’ll then construct a second mannequin with normalised inputs and illustrate its affect on the ultimate prediction end result.

Step one on this tutorial is to import the libraries we might be working with.

For this tutorial, we’d like 4 libraries:

These are imported as follows:

import pandas as pdfrom…