In order to anticipate failures and reduce downtime, “predictive diagnostic” aims not only at warning about the failure events before they occur but also at identifying the causes of degradation leading to such detections. Then, based on the results of predictive diagnostic, “prognostic” aims at estimating the remaining useful life in order to plan a maintenance action before unit performances are affected. However, these are complex tasks. To overcome these difficulties, the notion of fleet may be very useful. In the present paper a fleet is composed of heterogeneous units (mainly components but could be systems or sub-systems) that are grouped together considering some similarities. The fleet can provide capitalized data and information coming from other members of the fleet for the improvement/development of the diagnostic/prognostic models. In order to achieve PHM with a fleet-wide dimension, it is thus necessary to manage relevant knowledge arising from the fleet taking into account heterogeneities and similarities amongst components, operational context, behaviours, etc. This paper will focus mainly in the formalization of a data-driven prognostic model considering a fleet-wide approach. The model is based on a prognostic approach of the system health using Relevant Vector Machine. The proposed model is based on historical data coming from similar units of a fleet. The heterogeneity of the monitored data is treated by assessing a global health index of the units. The proposed approach is shown on a case study. This case study illustrates how the fleet dimension facilitates predictive diagnostic and the definition of the prognostic model in the marine domain.

Remaining Useful Life (RUL) of equipment is defined as the duration between the current time and its failure. An accurate and reliable prognostic of the remaining useful life provides decision-makers with valuable information to adopt an appropriate maintenance strategy to maximize equipment utilization and avoid costly breakdowns. In this work, we propose an end-to-end deep learning model based on dense layers and long short-term memory layers (LSTM) to predict the RUL. After normalization of all data, inputs are fed directly to several dense layers for feature learning, then to an LSTM layer to add the notion of time, and finally to other dense layers for RUL prediction. The proposed architecture is tested on the NASA commercial modular aero-propulsion system simulation (C-MAPSS) dataset. Despite its simplicity with respect to other recently proposed models, the model developed outperforms them with a significant decrease in the competition score and in the root mean square error between the predicted and the gold value of the RUL. In this paper, we will discuss how the proposed end-to-end model is able to achieve such good results and compare it to other deep learning and state-of-the-art methods.