Anomaly detection from IoT Time Series @AlfaLaval - Revision history

Islab at 23:44, 9 October 2021

2021-10-09T23:44:45Z

Islab at 23:44, 9 October 2021

2021-10-09T23:44:09Z

Islab: Anomaly detection from IoT Time Series @AlfaLaval

2021-09-29T13:01:48Z

Anomaly detection from IoT Time Series @AlfaLaval

New page

{{StudentProjectTemplate
|Summary=Anomaly detection from IoT Time Series @AlfaLaval
|Keywords=anomaly detection, time series, feature extraction, ensemble, recurrent neural networks, LSTM, lation forest
|References=(1) Anomaly detection: Aggarwal, Charu C., Outlier Analysis, 2017
(4) Christ, M., Kempa-Liehr, A.W. and Feindt, M. (2016). Distributed and parallel time series feature extraction for industrial big data applications. ArXiv e-prints: 1610.07717 URL: http://adsabs.harvard.edu/abs/2016arXiv161007717C
(4) Zhao, Y., Nasrullah, Z. and Li, Z., 2019. PyOD: A Python Toolbox for Scalable Outlier Detection. Journal of machine learning research (JMLR), 20(96), pp.1-7.
(5) Python Packages: keras, PyOD, tsfresh
(6) Tutorial:Time Series Analysis Lecture: https://www.youtube.com/watch?v=ZuydOEws92s

|Prerequisites=Data Mining Course: Lectures 5 (Time Series Analysis) and 6 (Anomaly Detection)

|Supervisor=Hadi Fanaee
|Level=Master
|Status=Open
}}
This is a fantastic opportunity to work with Alfa Laval, a world's leader and pioneer in producing separator machines. During the project, you will have this opportunity to gain access to real-life industrial IoT data and gain first-hand experience with such kind of valuable data.

This project aims to investigate the application of anomaly detection methods from multiple sensor time series from separator machines. The separators purify oil and water supplies onboard marine vessels.

It is not trivial to apply anomaly detection methods designed for tabular data to time series, due to special property of time series. The main objective of this project is to evaluate the usefulness of anomaly detection methods to industrial IoT time series in Alfa Laval AB, in particular investigation of two scenarios: feature extraction (e.g. tsfresh) + ensemble methods (e.g., isolation forest) vs. Recurrent Neural Networks (e.g., LSTM).

Hadi Fanaee, Assistant Professor Website: www.fanaee.com E-mail: hadi.fanaee@hh.se

@@ Line 3: / Line 3: @@
 |Keywords=anomaly detection, time series, feature extraction, ensemble, recurrent neural networks, LSTM, lation forest
 |References=(1) https://www.youtube.com/watch?v=ZuydOEws92s <br>
-(1) https://www.youtube.com/watch?v=sF2DeSPrGfc
+(2) https://www.youtube.com/watch?v=sF2DeSPrGfc
 |Prerequisites=Data Mining Course: Lectures 5 (Time Series Analysis) and 6 (Anomaly Detection)
 |Supervisor=Hadi Fanaee