This image shows Sheng Ding

Sheng Ding


Academic staff
Institute of Industrial Automation and Software Engineering


+49 711 685 67324

Pfaffenwaldring 47
70550 Stuttgart
Room: 2.301

Journals and conferences:
  1. 2021

    1. Y. Ma, A. Morozov, and S. Ding, “Anomaly Detection for Cyber Physical Systems using Transformers,” The International Mechanical Engineering Congress and Exposition, IMECE, 2021.
    2. S. Ding, A. Morozov, T. Fabarisov, and S. Vock, “KrakenBox: Deep Learning based Error Detector for Industrial Cyber-Physical Systems,” in The International Mechanical Engineering Congress and Exposition, IMECE, 2021.
  2. 2020

    1. S. Ding, A. Morozov, S. Vock, M. Weyrich, and K. Janschek, “Model-Based Error Detection for Industrial Automation Systems Using LSTM Networks,” In: International Symposium on Model-Based Safety and Assessment, pp. 212-226. Springer, Cham, 2020.
  3. 2019

    1. K. Ding, S. Ding, A. Morozov, T. Fabarisov, and K. Janschek, “On-line error detection and mitigation for time-series data of cyber-physical systems using deep learning based methods,” in 2019 15th European Dependable Computing Conference (EDCC) (pp. 7-14). IEEE, 2019.

Sheng Ding is a PhD Student and Research Associate at the University of Stuttgart, Institute of Automation Technology and Software Engineering. His main research interests are deep learning based error detection and reliability analysis of cyber-physical systems.

Research focus:

Deep learning based error detection and reliability analysis of cyber-physical systems.


Anomaly detection is a well-known concept that is used in a wide variety of fields, including systems engineering, to detect or prevent errors. As deep learning (DL) approaches continue to improve, the applicability of DL techniques for error detection is being explored. This research will work to focus on time series computing and common CPS errors. First, the training data is generated using a combination of existing and own error injection methods and tools. For this purpose, data are generated using simulation in the area of networked automation systems and errors through injection with the appropriate label. So that the methods can be implemented in the real world, real case studies are considered and compared measured signals with the simulation.


The vast majority of existing research on anomaly detection does not consider real-time. However, this is very important to protect a CPS system. On the one hand, online real-time measurement and data traffic from the host system should be documented and evaluated. On the other hand, online real-time response is explored by comparing the computation costs of different DL models.

In the project, many types of methods or DL architectures are compared and integrated. Appropriate DL-based error detection approaches such as prediction, classification, and reconstruction are compared. The training and evaluation of the DL models take place offline with the collected data sets and comprise three phases. First, different network types, e.g. MLP, CNN, LSTM / GRU, Autoen-Coders, and transformers, as well as their structure and hyper-parameters are examined and identified. Second, the performance with accuracy and speed of various models is documented as a knowledge base. Third, to improve accuracy, dynamic reconfiguration is being researched. The aim is to develop a DL-based error detection with real-time reaction and high accuracy. The implemented method is tested online to demonstrate the concept. The results show that the errors can be successfully identified in real-time.

Research portal:

Google Scholar:



To the top of the page