Optimizing maintenance is essential for industrials to stay competitive, and the development of appropriate predictive maintenance is necessary to achieve this objective. To this extent, the Prognostics and Health Management (PHM) paradigm is well established. One of the key steps of PHM is the prognostics of health states of the system. Various state-of-the-art approaches exist for prognostics, with an emerging orientation towards data-driven methods. Indeed, they have lot of potential for Industry 4.0 applications with high amount of data from sensors and control equipment. However, labelled data (i.e., failures of systems) is not always available on real-life applications where preventive maintenance is often already applied. Thus, the learning databases can be unbalanced, with few learning examples, consequently reducing the learning capacities of algorithms, as well as their generalization. One way to optimize learning on such applications is then to use Expert Knowledge, which can provide additional information on the system and its operating model. A challenging issue is herein the development of a general methodology to integrate the Expert Knowledge into data-driven methods.To face this challenge, this paper aims to propose a categorization of Expert Knowledge based on existing works to identify adapted methods that can help to integrate efficiently the available Knowledge into relevant prognostics algorithms. The proposed categorization will allow and facilitate the review and comparison of approaches and methodologies introduced in the literature and in further research. Finally, the proposed classification will be illustrated on a real case of prognostics for a hydraulic circuit from an ArcelorMittal plant.

The purpose of this work is to evaluate the performance of several wear models, either with different mathematical formulation or different definition of the unknown wear coefficients, on the prediction of the work-roll wear amplitude in Hot Strip Mills (HSM). To achieve this goal, a classical model calibration approach based on inverse optimization has been developed to calibrate these several wear models. A large industrial hot rolling database composed by roll wear amplitude measurements for both later finishing mill stands (F6 and F7) from ArcelorMittal Dofasco HSM was considered and a least-square cost function was applied to minimize the differences between both numerical and experimental results during the optimization process. The averaged roll wear gap between measurements and optimized numerical predictions was then used as a quantitative indicator to compare the performance between the wear models and identify the most suitable one for roll wear prediction. In addition, an Artificial Neural Network (ANN) approach was developed based on the most suitable wear model. Thus, roll wear predictions obtained using the ANN were compared with the ones obtained using Classical calibration to evaluate the performance of both approaches.

No abstract