Predictive Maintenance (PdM) is the newest strategy for maintenance management in industrial contexts. It aims to predict the occurrence of a failure to minimize unexpected downtimes and maximize the useful life of components. In data-driven approaches, PdM makes use of Machine Learning (ML) algorithms to extract relevant features from signals, identify and classify possible faults (diagnostics), and predict the components’ remaining useful life (prognostics). The major challenge lies in the high complexity of industrial plants, where both operational conditions change over time and a large number of unknown modes occur. A solution to this problem is offered by novelty detection, where a representation of the machinery normal operating state is learned and compared with online measurements to identify new operating conditions. In this paper, a systematic study of autoencoder-based methods for novelty detection is conducted. We introduce an architecture template, which includes a classification layer to detect and separate the operative conditions, and a localizer for identifying the most influencing signals. Four implementations, with different deep learning models, are described and used to evaluate the approach on data collected from a test rig. The evaluation shows the effectiveness of the architecture and that the autoencoders outperform the current baselines.

In the past decade, many approaches have been suggested to execute ML workloads on a DBMS. However, most of them have looked at in-DBMS ML from a training perspective, whereas ML inference has been largely overlooked. We think that this is an important gap to fill for two main reasons: (1) in the near future, every application will be infused with some sort of ML capability; (2) behind every web page, application, and enterprise there is a DBMS, whereby in-DBMS inference is an appealing solution both for efficiency (e.g., less data movement), performance (e.g., cross-optimizations between relational operators and ML) and governance. In this paper, we study whether DBMSs are a good fit for prediction serving. We introduce a technique for translating trained ML pipelines containing both featurizers (e.g., one-hot encoding) and models (e.g., linear and tree-based models) into SQL queries, and we compare in-DBMS performance against popular ML frameworks such as Sklearn and ml.net . Our experiments show that, when pushed inside a DBMS, trained ML pipelines can have performance comparable to ML frameworks in several scenarios, while they perform quite poorly on text featurization and over (even simple) neural networks.