ABGP - I Workshop AI applied to The Petroleum Industry
The use of neural networks and especially the convolutional ones are increasing dramatically, as evidenced by the number of scientific articles published in recent years about the subject, including in geosciences.
Convolutional neural networks are widely used for image recognition and classification. From a large range of examples, the algorithms are trained to recognize certain patterns. The CNNs pre-processing includes segmenting the input large- scale image into thousands of randomly selected smaller squares images and, for each sample, the algorithm performs convolutions with various filters to recognize images features. Through the analysis of accuracy and loss function, after some iterations, the algorithm is able to recognize textures and classify patterns. The key to the success of the methodology is to develop an algorithm or workflow that produces a concrete, robust and efficient result. In the present work some tests were made using a CNN developed by the authors for mapping the salt section in 2D pre-salt lines in the Santos basin, in order to test the sensitivity of the algorithm to the different input parameters.
