(128b) Anomaly Detection in Process Data Using Generative Adversarial Networks (GAN)

The development of DetlaV Advanced Continuous Historian has paved the way for highly scalable storage of live plant data in the time series format. Conducting anomaly detection in such abundant process data can provide insights into irregular plant behavior and advise process engineers to take proactive measures. The recent proliferation of deep learning methods which excel in learning complex spatial/temporal correlations shows great promises in anomaly detection. In current work, we demonstrate the adoption of an unsupervised times series anomaly detection framework based on generative adversarial networks (GAN) [1, 2]. The efficacy and efficiency of the framework is illustrated using both open source data and industrial data. Such results demonstrate the successful application of state-of-art artificial intelligence to cope with real problems in chemical manufacturing within the Digital Transformation era.

[1] Geiger, A., Liu, D., Alnegheimish, S., Cuesta-Infante, A., & Veeramachaneni, K. (2020). TadGAN: Time Series Anomaly Detection Using Generative Adversarial Networks. arXiv preprint arXiv:2009.07769.

[2] Li, D., Chen, D., Jin, B., Shi, L., Goh, J., & Ng, S. K. (2019, September). MAD-GAN: Multivariate anomaly detection for time series data with generative adversarial networks. In International Conference on Artificial Neural Networks (pp. 703-716). Springer, Cham.