An example of bulk density prediction (RHOB) with Keras and illustrates the impacts of normalization on the prediction results.
Large amounts of data are acquired daily from wells around the world. However, the quality of that data can vary significantly from missing data to data affected by sensor failures and well conditions. This can have consequences on other parts of an underground project, such as delays and inaccurate assumptions and conclusions.
As missing data is one of the most common problems we face with well log data quality, numerous methods and techniques have been developed to estimate the values and fill in the gaps. This includes the application of machine learning technology, which has increased in popularity in recent decades with libraries such as TensorFlow and PyTorch.
In this tutorial, we will use Keras, which is a high-level neural network API that runs on top of TensorFlow. We will use it to illustrate the process of building a machine learning model to enable bulk density (RHOB) predictions. This is a commonly acquired recording measurement; However, it can be significantly affected by poor well conditions or, in some cases, tools can fail, resulting in measurements not being made at key intervals.
We’ll start with a very simple model, which does not take into account normalization of inputs, a common step in the machine learning workflow. Then, we will build a second model with normalized inputs and illustrate its impact on the final prediction result.
The first step in this tutorial is to import the libraries we will work with.
For this tutorial, we need 4 libraries:
These are imported as follows:
import pandas as pd
from…
