Constraint Guided AutoEncoders for Joint Optimization of Condition Indicator Estimation and Anomaly Detection in Machine Condition Monitoring

TL;DR

An extension to Constraint Guided AutoEncoders (CGAE), which is a robust AD method, is proposed that enables building a single model that can estimate a CI that shows monotonic behavior over time, reflecting the expected gradual degradation of the asset’s condition.

Abstract

The main goal of machine condition monitoring is, as the name implies, to monitor the condition of industrial applications. The objective of this monitoring can be mainly split into two problems. A diagnostic problem, where normal data should be distinguished from anomalous data, otherwise called Anomaly Detection (AD), or a prognostic problem, where the aim is to predict the evolution of a Condition Indicator (CI) that reflects the condition of an asset throughout its life time. When considering machine condition monitoring, it is expected that this CI shows a monotonic behavior, as the condition of a machine gradually degrades over time. This work proposes an extension to Constraint Guided AutoEncoders (CGAE), which is a robust AD method, that enables building a single model that can be used for both AD and CI estimation. For the purpose of improved CI estimation the extension incorporates a constraint that enforces the model to have monotonically increasing CI predictions over time. Experimental results indicate that the proposed algorithm performs similar, or slightly better, than CGAE, with regards to AD, while improving the monotonic behavior of the CI.

Figures (7)

• Figure 1: Example of a bearing test rig setup.
• Figure 2: The microphone setup used in the bearing run-to-failure tests.
• Figure 3: The mean (a) and standard deviation (c) of the BA on the SM dataset, and the mean (b) and standard deviation (d) on the ABM dataset, for the different methods and different thresholds. The thresholds are $T_{train}$, $T_{sigmoid}$, and $T_{opt}$, as described in Section \ref{['ssec:metrics']}. The x-axis indicates from how many runs the anomalous data is used.
• Figure 4: The mean (a) and standard deviation (c) of the BA on the SM dataset, and the mean (b) and standard deviation (d) on the ABM dataset, for the different reconstructions. The x-axis indicates the number runs from which anomalous data is used.
• Figure 5: The mean (a) and standard deviation (c) of the relative difference in norm between the different thresholds and the corresponding optimal threshold for the BA on the SM dataset, as well as the ABM dataset (b) and (d), for the different methods. The thresholds are $T_{train}$, $T_{sigmoid}$, as described in Section \ref{['ssec:ae-dsvdd']}.