No abstract

No abstract

Helicopter is a system on which operates a large number of specialities from electrical domain to mechanical domain. Today, diagnosis methods are segmented by field of expertise. Each expert treats the sub-system whose he is responsible for regardless of the results of other specialities.Therefore, there is no relation between specialities. Thus, diagnosis at system level is efficient but, due to the lack of correlation between subsystems, it is incomplete. Within the framework of a helicopter, the operator of maintenance uses all the data recorded during the flight, the results of expert treatments, but also his knowledge, his experience and his capacities of observation and analysis to provide an effective global diagnosis. In order to build relation between fields of expertise and so, to obtain a diagnosis at aircraft level which could be relevant, we try to set up a concept which gets closer, at most, to human judgment but which is not well adapted to industrial environment: the normality.During our study, we established that to reach our objective and to stick at best with the concept of normality, the most relevant solution consists in building a global normal signature. This signature could be illustrated as the image of the aircraft health, qualified as normal. This paper defines the normal signature and explains, in part, the process of its building.

For helicopters, anticipating failures is crucial. To this end, the analysis of flight data allows to develop predictive maintenance approaches, for which Airbus Helicopters (AH) has proposed several solutions, some based on machine learning using predictive models. One recurrent problem in this setting is the contextualization of the data, that is to identify the data better fitting the phenomenon being modeled. Indeed, helicopters are complex systems going through different flight phases. Experts therefore have to identify the adequate ones, in which the selected flight parameters are stable and consistent with the studied problem. In this paper, we propose a generic solution to contextualize classification data, and present an experimental study on AH flight data: the results are encouraging and allow to keep domain experts involved the process.

Huge amount of multivariate time series (TS) data are recorded by helicopters in operation, such as oil temperature, oil pressure, altitude, rotor speed to mention a few. Despite the effort deployed by Airbus Helicopters towards an effective use of those TS data, getting meaningful and intuitive representations of them is a never ending process, especially for domain experts who have a limited time budget to get the main insights delivered by data scientists. In this paper, we introduce a simple yet powerful and scalable technique for visualizing large amount of TS data through patterns movies. We borrow the co-occurrence matrix concept from image processing, to create 2D pictures, seen as patterns, from any multivariate TS according to two dimensions over a given period of time. The cascade of such patterns over time produces so-called patterns movies, offering in a few seconds a visualisation of helicopter' parameters in operation over a long period of time, typically one year. We have implemented and conducted experiments on Airbus Helicopters flight data. First outcomes of domain experts on patterns movies are presented.