subsurfaceAI

Module 3

Request a Demo
Request a Demo

Module 3: Well Log Data Analysis and Prediction

Module 3 focuses on enhancing well log data analysis and prediction capabilities. It offers advanced noise filtering and data conditioning techniques to remove noise from log data effectively. The module includes an elastic modulus calculator for generating essential log curves based on Vp, Vs, and density logs, adaptable for individual or group wells. It supports comprehensive data analysis to estimate log distribution and correlation across facies. The module enables 2-D and 3-D cross plotting of well logs, real-time updates of samples in well sections, and implements four innovative methods for log facies classification, including deterministic rules, self-organizing maps, and supervised classifications through deep-learning and Bayes models. Additionally, it provides tools for predicting missing logs and analyzing the performance of prediction models to prevent over-training, marking a significant advancement in well log data interpretation and analysis.

Key Features

  • Noise Filtering and Data Conditioning:
    • Remove noise in log data through noise filtering and de-spiking techniques.
  • Elastic Modulus Calculator:
    • Calculate 12 elastic modulus log curves from Vp, Vs, and density logs for a single well or a group of wells. The flexible log calculator can derive user-defined log curves.
  • Data Analysis:
    • Estimate log distribution and correlation for each facies.
  • 2-D and 3-D Cross Plotting:
    • Flexibly select any well logs to cross-plot in 2-D and 3-D, color-coded by any well log.
    • Update samples in user-defined polygons on cross-plots in real-time.
  • Log Facies Classification:
    • Four methods implemented for generating facies logs:
      1. Deterministic rules from one or multiple log curves.
      2. Self-organizing map from one or multiple logs.
      3. Supervised classification based on deep-learning models and the Bayes model.
  • Predict Missing Logs:
    • Build deep learning models to predict missing logs from other logs.

Perform performance analysis of the prediction model to avoid over-training.