Abstract
This paper presents a novel approach for anomaly detection in industrial processes. The system solely relies on unlabeled data and employs a 1D-convolutional neural network-based deep autoencoder architecture. As a core novelty, we split the autoencoder latent space in discriminative and reconstructive latent features and introduce an auxiliary loss based on k-means clustering for the discriminatory latent variables. We employ a Top-K clustering objective for separating the latent space, selecting the most discriminative features from the latent space. We use the approach to the benchmark Tennessee Eastman data set to prove its applicability. We provide different ablation studies and analyze the method concerning various downstream tasks, including anomaly detection, binary and multi-class classification. The obtained results show the potential of the approach to improve downstream tasks compared to standard autoencoder architectures.
Highlights
IntroductionIntelligent condition monitoring of such processes is necessary to analyze changes in the process parameters and determine anomalies that hurt the reliability of the overall system
We present a novel unsupervised learning approach based on 1-dimensional convolutional neural networks and deep autoencoder structure where we define an auxiliary loss to increase the expressiveness of the latent representation
We presented a novel approach for unsupervised training of time series data sets the proposed Top-K DCCA model with other existing approaches
Summary
Intelligent condition monitoring of such processes is necessary to analyze changes in the process parameters and determine anomalies that hurt the reliability of the overall system. A significant part of data-driven methods, namely supervised machine learning relies on the availability of labelled data from all of the possible operating conditions of the system. This availability of labelled data for industrial processes is infeasible due to various reasons. Data sets for inconceivable fault cases are impossible to gather In such cases, unsupervised or semi-supervised learning based data-driven techniques is the only alternative as they can suitably characterize the fault-free state of the system, which can subsequently be used to assess abnormal or faulty conditions.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have