Including Sparse Production Knowledge into Variational Autoencoders to Increase Anomaly Detection Reliability

TL;DR

This paper introduces a novel approach that incorporates sparse production knowledge into Variational Autoencoders, significantly enhancing anomaly detection accuracy in manufacturing data.

Contribution

It presents a method to integrate infrequent labeled data into VAEs, improving anomaly detection beyond traditional unsupervised and supervised models.

Findings

Outperforms PCA, Isolation Forest, and neural networks in accuracy

Achieves higher precision and recall in anomaly detection

Effective use of sparse labels boosts detection reliability

Abstract

Digitalization leads to data transparency for production systems that we can benefit from with data-driven analysis methods like neural networks. For example, automated anomaly detection enables saving resources and optimizing the production. We study using rarely occurring information about labeled anomalies into Variational Autoencoder neural network structures to overcome information deficits of supervised and unsupervised approaches. This method outperforms all other models in terms of accuracy, precision, and recall. We evaluate the following methods: Principal Component Analysis, Isolation Forest, Classifying Neural Networks, and Variational Autoencoders on seven time series datasets to find the best performing detection methods. We extend this idea to include more infrequently occurring meta information about production processes. This use of sparse labels, both of anomalies or production data, allows to harness any additional information available for increasing anomaly detection performance.