E&PRiskIV
Description: During the development of the Neural Module of E&PRiskIII, it was observed that a “critical” point for a good performance of the Neural Module with respect to the capacity of forecast of total time of operation concerns the quality of the historical data that are used for neural network training and found in a database provided by the user. If the training set is not sufficiently representative with respect to the various conditions and alternatives of drilling/completion operation parameters, the neural network will have its generalization capacity extremely impaired resulting in inaccurate or even incorrect predictions. The aim is to research and develop intelligent and automated data mining techniques applied to databases of drilling/completion operations in order to decrease or withdraw from the user the responsibility for the quality of the data to be used for neural network training. Thus, this module aims to propose an automated process of data mining in databases of drilling/completion for the use of neural networks as knowledge extractor and prediction of operating times.
Status: Completed.
Nature: Development.
Students involved: Undergraduate: (2).
Members: Mauro Roisenberg – Member / Silvia Modesto Nassar – Member / Dalton Francisco de Andrade – Member / Dyego Wuebel Santin – Member / Leonardo Freitas Noleto – Member / Paulo José de Freitas Filho – Coordinator.
Financier(s): Petróleo Brasileiro – Rio de Janeiro – Matriz – Financial aid.
Years: 2005-2007.
