Novelty Detection with Autoencoders for System Health Monitoring in Industrial Environments

Abstract

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.

A Real-Time Novelty Recognition Framework Based on Machine Learning for Fault Detection

“…[16] or Ref. [17]. Generally, the Supervised/Unsupervised choice depends on how much and which type of data has been collected.…”

“…Papers such as Refs. [5], [17], or [25] have worked on this topic, framing this as a new research field to help companies implement a preliminary framework to collect data until a real PdM infrastructure is built. This detection type can be applied in numerous domains such as medical, robotic/industrial, vision surveillance, and other fields, as reported in Ref.…”

PrOuD: Probabilistic Outlier Detection Solution for Time-Series Analysis of Real-World Photovoltaic Inverters

“…The authors of [9,10] proposed unsupervised anomaly detection methods using a convolutional approach coupled to an autoencoder framework. In addition, hybrid methods combining a convolutional neural network (CNN) and long short-term memory (LSTM) have been increasingly proposed, such as [10][11][12][13]. A graph neural network (GNN) considers correlations among sensors and hidden relationships within multivariate time series, and it has also found application in the domain of anomaly detection [14][15][16].…”

Incremental Learning of Human Activities in Smart Homes

“…This will result in some form of novelty score, which is then compared with a decision threshold, where new unseen inputs are classified as novel if the threshold is exceeded. Novelty detection has gained much research attention, especially in diagnostic and monitoring systems [ 10 , 11 , 12 ]. An overview of the existing approaches is provided in [ 13 ].…”