Data-driven drift detection and diagnostic for heterogeneous production process
Résumé
The access to pertinent data from the implementation of new Industry 4.0 technologies has challenged and ameliorated many areas of the industry. In that way, industrial maintenance area has seen some of its deployment strategies greatly improved by data-driven deployment approaches. It is truer for predictive maintenance strategy. Indeed, for example, data-based prognostics allows calculating a RUL closer to the real degradation of the system to be maintained. However, not all the predictive maintenance processes are well addressed by the scientific community. More precisely, those related to drift detection and drift diagnosis of heterogeneous production processes are few investigated. To face this issue, the scope of this paper is to propose a methodology based on Artificial Intelligence techniques to support the development of these two processes. Then, the methodology is validated on two heterogeneous production processes of a French SEW-USOCOME factory supplying electrical motors.