Predictive maintenance for offshore oil wells by means of deep learning features extraction

Nowadays, the great diffusion of Internet Things and improvements in Artificial Intelligence techniques have given a rise development application data-driven approaches for Predictive Maintenance to reduce costs linked maintenance industrial machinery. Due wide real-life applications strong interest by even more industries, this field is highly attractive academics practitioners. So, constructing efficient frameworks address problem an open debate. In work, we propose Deep Learning approach feature extraction offshore oil wells monitoring context, exploiting public 3 W dataset, which well-known literature. The dataset made up about 2000 multivariate time series labelled according corresponding functioning well. there classification task with eight classes, each related particular machinery condition. Thanks peculiarities labels, proposed framework valid both diagnostics prognostics. detail, compare two different extraction. first statistical approach, widely used literature considered dataset; second based on Convolutional 1D AutoEncoder. extracted features are then as input several Machine algorithms, namely Random Forest, Nearest Neighbours, Gaussian Naive Bayes Quadratic Discriminant Analysis. Different experiments various horizons prove worthiness