EMERGE

Emerge is a geostatistical, attribute prediction module that can predict property volumes using well logs and attributes from seismic data. The predicted properties can be any log types available, such as porosity, velocity, density, gamma-ray, lithology and water saturation. Emerge can also be used to predict missing logs or parts of logs by using existing logs that are common to the available wells.

Multi-Attribute Analysis

Using multi-linear regression or neural network analysis, Emerge trains itself at the well locations to learn the optimum transform that relates the logs and seismic data. It then applies that transform to derive a volume of the log property from the seismic volume(s).

Inversion Benefits with Emerge

• Ability to predict volumes of any data type
• Internal cross-validation allows all data to be used during training
• Derives a measure of correlation and error at each well used in the training
• Enhanced prediction of missing log data by using multi-log and non-linear combinations
• Non-linear, high-resolution predictions using Neural Networks

Attribute Prediction Features

• Filter and normalize logs
• Cross-plot logs and seismic data
• Perform non-linear transforms on logs and seismic data
• Compute seismic attributes internally
• Compute Principal Components
• Utilize an unlimited number of external seismic volumes
• Predict volumes of any log type (recorded or computed)
• Perform stepwise multi-linear regression for the rapid selection of useful attributes
• Optional convolutional operator to extend multi-linear regressions
• Automatically validate predictions to avoid over-training
• User-defined blind well testing

Emerge log predict uses the same multi-attribute methodology as seismic attribute prediction, but applies it to log data. It can predict missing logs or parts of logs not by using seismic data, but by using existing logs that are common to the wells in the training data set.

Emerge Prediction Options

• Single-Attribute Regression
• Multi-Attribute Regression
• Probabilistic Neural Network
• Multi-Layer Feed Forward Neural Network
• Radial Basis Function Neural Network