In the last five years, the inclusion of Deep Learning algorithms in prognostics and health management (PHM) has led to a performance increase in diagnostics, prognostics, and anomaly detection. However, the lack of interpretability of these models results in resistance towards their deployment. Deep Learning-based models fall within the accuracy/interpretability tradeoff, which means that their complexity leads to high performance levels but lacks interpretability. This work aims at addressing this tradeoff by proposing a technique for feature selection embedded in deep neural networks that uses a feature selection (FS) layer trained with the rest of the network to evaluate the input features’ importance. The importance values are used to determine which will be considered for deployment of a PHM model. For comparison with other techniques, this paper introduces a new metric called ranking quality score (RQS), that measures how performance evolves while following the corresponding ranking. The proposed framework is exemplified with three case studies involving health state diagnostics and prognostics and remaining useful life prediction. Results show that the proposed technique achieves higher RQS than the compared techniques, while maintaining the same performance level when compared to the same model but without an FS layer.

Due to its capital-intensive nature, the Oil and Gas industry requires high operational standards to meet safety and environmental requirements, while maintaining economical returns. In this context, maintenance policies play a crucial role in the avoidance of unplanned downtimes and enhancement of productivity. In particular, Condition-Based Maintenance is an approach in which maintenance actions are performed depending on the assets’ health state that is evaluated through different kinds of sensors. In this paper, Deep Learning methods are explored and different models are proposed for health state prognostics of physical assets in two real-life cases from the Oil and Gas industry: a Natural Gas treatment plant in an offshore production platform where elevated levels of CO₂ must be predicted, and a sea water injection pump for oil extraction stimulation, in which several degradation levels must be predicted. A general methodology for preprocessing the available multi-sensor data and developing proper models is proposed and apply in both case studies. In the first one, a LSTM autoencoder is developed, achieving precision values over 83.5% when predicting anomalous states up to 8 h ahead. In the second case study, a CNN-LSTM model is proposed for the pump’s health state prognostics 48 h ahead, achieving precision values above 99% for all possible pump health states.