Modeling Mixed-Type Time Series Data With Neural Networks for Predictive Maintenance

Master's Thesis

Modeling Mixed-Type Time Series Data With Neural Networks for Predictive Maintenance


An Ngyuen (M.Sc.), Prof. Dr. B. Eskofier




Predictive Maintenance (PdM) allows companies to monitor operating conditions of systems to
schedule maintenance activities based on an as-needed basis. This can prevent unexpected equipment
downtime due to equipment failure as well as too early repairs [1].
Most work on PdM so far focuses on uni- or multivariate and regularly spaced sensor measurements
[2]. These datasets are mostly collected in a controlled experimental environment [3, 4, 5].
Other work focuses on event log data or frequent patterns derived from event log data [6, 7].
However, a complete picture about the condition of modern systems may only be obtained by
integrating different data sources like event logs and sensor measurements [8, 9].
The goal of this thesis is to develop, implement and evaluate neural network-based algorithms to
predict the breakdown of X-Ray tubes installed in CT Scanners. For this purpose, time series data
of a large fleet of CT scanners will be analyzed. The data includes event logs, irregularly spaced
physical sensor measurements, and usage statistics. In summary, the available data imposes the
following challenges:
– Different data sources need to be integrated
– Time series data is irregularly sampled
– Varying utilization of CT scanners between healthcare facilities
– Collected data is noisy since it is collected from real healthcare facilities
There already exist solutions from other domains, like [10, 11, 12], to tackle individual points
of the listed challenges. Therefore, this work aims at integrating and extending different existing
solutions to tackle all of the challenges that come with the specific data at hand.




[1] Mobley, R. Keith : An Introduction to Predictive Maintenance. Elsevier Science. 2002.
[2] Zhao, Rui et al.: Deep learning and its applications to machine health monitoring Mechanical
Systems and Signal Processing. 2019.
[3] Hai Qiu et al.: Wavelet Filter-based Weak Signature Detection Method and its Application
on Roller Bearing Prognostics. Journal of Sound and Vibration 289, 1066-1090. 2006.
[4] Smith, Wade A. et al.: Rolling element bearing diagnostics using the Case Western Reserve
University data: A benchmark study. Mechanical Systems and Signal Processing 64, 100-
131. 2015.
[5] Saxena, Abhinav, et al.: Damage propagation modeling for aircraft engine run-to-failure
simulation. International Conference on Prognostics and Health Management. 2008.
[6] Sipos, Ruben, et al.: Log-based predictive maintenance. Proceedings of the 20th ACM
SIGKDD International Conference on Dnowledge Discovery and Data Mining. 2014.
[7] Wang, J., et al.: Predictive maintenance based on event-log analysis: A case study. IBM
Journal of Research and Development 61.1, 11:121 – 11:132. 2017.
[8] Diez-Olivan et al.: Data fusion and machine learning for industrial prognosis: Trends and
perspectives towards Industry 4.0 Information Fusion. 2019.
[9] Feremans et al.: Pattern-Based Anomaly Detection in Mixed-Type Time Series Joint European
Conference on Machine Learning and Knowledge Discovery in Databases. 2019.
[10] Xiao, Shuai, et al.: Modeling the Intensity Function of Point Process via Recurrent Neural
Networks . Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence 5.
[11] Baytas, Inci M., et al: Patient Subtyping via Time-Aware LSTM Networks. KDD ’17:
Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery
and Data Mining, 65 – 74. 2017.
[12] Neil, Daniel et al.: Phased lstm: Accelerating recurrent network training for long or eventbased
sequences. Advances in Neural Information Processing Systems. 2016.
[13] Hochreiter, Sepp, and JÃ1
4rgen Schmidhuber: Long short-term memory. Neural computation
9.8. 1997.
[14] Cho, Kyunghyun, et al: Learning phrase representations using RNN encoder-decoder for
statistical machine translation. arXiv preprint. 2014.
[15] Ribeiro et al.: Why should i trust you?: Explaining the predictions of any classifier. Proceedings
of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and
Data Mining. 2016.
[16] Lundberg, Scott M., and Su-In Lee.: A unified approach to interpreting model predictions.
Advances in Neural Information Processing Systems. 2017