Anomaly Detection in Asset Degradation Process Using Variational Autoencoder and Explanations

Development of predictive maintenance (PdM) solutions is one of the key aspects of Industry 4.0. In recent years, more attention has been paid to data-driven techniques, which use machine learning to monitor the health of an industrial asset. The major issue in the implementation of PdM models is a lack of good quality labelled data. In the paper we present how unsupervised learning using a variational autoencoder may be used to monitor the wear of rolls in a hot strip mill, a part of a steel-making site. As an additional benchmark we use a simulated turbofan engine data set provided by NASA. We also use explainability methods in order to understand the model’s predictions. The results show that the variational autoencoder slightly outperforms the base autoencoder architecture in anomaly detection tasks. However, its performance on the real use-case does not make it a production-ready solution for industry and should be a matter of further research. Furthermore, the information obtained from the explainability model can increase the reliability of the proposed artificial intelligence-based solution.

Improvement in the Resistance to Wear of Work-Rolls Used in Finishing Stands of the Hot Strip Mills

Work-rolls manufactured through the Indefinite Chill Double Poured (ICDP) method present an exterior work layer manufactured in a martensitic white cast iron alloyed with 4.5 %Ni, 1.7 %Cr, and 0.7 %Nb (wt.%). In its microstructure, there are abundant carbides of the type M3C and MC, which give high resistance to wear, and graphite particles which improve the service behaviour of the rolls against thermal cycling. The core of the rolls is manufactured in grey cast iron of pearlitic matrix and spheroidal graphite. These work-rolls are used in the finishing stands in Hot Strip Mills for rolling slabs proceeding from continuous casting at 1200 °C. Through the application of a Design of Experiments (DoE), an attempt has been made to identify those manufacturing factors which have a significant effect on resistance to wear of these rolls and to find an optimal combination of levels of these factors which allow for improvement in resistance to wear. To increase resistance to wear, it is recommended to situate, simultaneously, the liquidus temperature and the percentage of Si in the respective ranges of 1250–1255 °C and 1.1–1.15 (wt.%). Higher liquidus temperatures favour the presence of the pro-eutectic constituent rather than the eutectic constituent. The outer zone of the work layer, in contact with the metal sheet, which is being rolled, does not show the graphitising effect of Si (0.8–1.15 wt.%). On the contrary, it confirms the hardening effect of the Si in solid solution of the ferrite. The addition of 0.02% of Mg (wt.%) and the inoculation of 6 kg/T of FeB tend to eliminate the graphitising effect of the Si, thus favouring that the undissolved carbon in the austenite is found to form carbides in contrast to the majority formation of graphite.