Position innerhalb des Seitenbaumes

Institut für Automatisierungstechnik und Softwaresysteme
Institut
Team
Ma, Yuliang

Yuliang Ma

M.Sc.

Akademischer Mitarbeiter
Institut für Automatisierungstechnik und Softwaresysteme

Kontakt

+49 711 685 69198
+49 711 685 67302
E-Mail

Pfaffenwaldring 47
70550 Stuttgart
Raum: 3.252

Publikationen

Zeitschriften und Konferenzen:

2022
1. P. Grimmeisen, M. Diaconeasa, Y. Ma, und A. Morozov, „Automated Generation of Hybrid Probabilistic Risk Models from SysML v2 Models of Software-Defined Manufacturing Systems“, in Morozov ASME 2022 International Mechanical Engineering Congress & Exposition (IMECE 2022), 30.Oct-3. Nov. 2022, 2022, 2022.
  - BibTeX
  BibTeX
  @conference{grimmeisen2022automated, author = {Grimmeisen, Philipp and Diaconeasa, Mihai and Ma, Yuliang and Morozov, Andrey}, booktitle = {Morozov ASME 2022 International Mechanical Engineering Congress & Exposition (IMECE 2022), 30.Oct-3. Nov. 2022, 2022}, title = {Automated Generation of Hybrid Probabilistic Risk Models from SysML v2 Models of Software-Defined Manufacturing Systems}, year = 2022 }
2. Y. Ma, P. Grimmeisen, und A. Morozov, „Detection and classification of robotic manipulator anomalies using MLSTM-FCN models“, ASME 2022 International Mechanical Engineering Congress & Exposition (IMECE 2022), 30.Oct-3. Nov. 2022, 2022, 2022.
  - BibTeX
  BibTeX
  @article{noauthororeditor2022detection, author = {Ma, Yuliang and Grimmeisen, Philipp and Morozov, Andrey}, journal = {ASME 2022 International Mechanical Engineering Congress & Exposition (IMECE 2022), 30.Oct-3. Nov. 2022, 2022}, title = {Detection and classification of robotic manipulator anomalies using MLSTM-FCN models}, year = 2022 }
2021
1. Y. Ma, A. Morozov, und S. Ding, „Anomaly Detection for Cyber Physical Systems using Transformers“, The International Mechanical Engineering Congress and Exposition, IMECE, 2021.
  - Zusammenfassung
  - BibTeX
  Zusammenfassung
  Safety and reliability are two critical factors of modern Cyber-Physical Systems (CPS). However, the increasing structural and behavioral complexity of modern automation systems significantly increases the possibility of system errors and failures, which can easily lead to economic loss or even hazardous events. Anomaly Detection (AD) techniques provide a potential solution to this problem, and conventional methods, e.g., Autoregressive Integrated Moving Average model (ARIMA), are no longer the best choice for anomaly detection for modern complex CPS. Recently, Deep Learning (DL) and Machine Learning (ML) anomaly detection methods became more popular, and numerous practical applications have been presented in many industrial scenarios. Most of the modern DL-based anomaly detection methods use the prediction approach and LSTM architecture. The Transformer is a new neural network architecture that outperforms LSTM in natural language processing. In this paper, we show that the Transformer-based deep learning model, which has received much attention recently, can be applied to the anomaly detection of industrial automation systems. Specifically, we collect time-series data from a system of two industrial robots using a Simulink model. Then, we feed these data into our Transformer-based model and train it to be a time-series data predictor. The paper presents the experimental results that show the comparison of precision and speed of a Long-Short Time Memory (LSTM) predictor and our Transformer-based predictor.
  BibTeX
  @article{ma2021anomaly, abstract = {Safety and reliability are two critical factors of modern Cyber-Physical Systems (CPS). However, the increasing structural and behavioral complexity of modern automation systems significantly increases the possibility of system errors and failures, which can easily lead to economic loss or even hazardous events. Anomaly Detection (AD) techniques provide a potential solution to this problem, and conventional methods, e.g., Autoregressive Integrated Moving Average model (ARIMA), are no longer the best choice for anomaly detection for modern complex CPS. Recently, Deep Learning (DL) and Machine Learning (ML) anomaly detection methods became more popular, and numerous practical applications have been presented in many industrial scenarios. Most of the modern DL-based anomaly detection methods use the prediction approach and LSTM architecture. The Transformer is a new neural network architecture that outperforms LSTM in natural language processing. In this paper, we show that the Transformer-based deep learning model, which has received much attention recently, can be applied to the anomaly detection of industrial automation systems. Specifically, we collect time-series data from a system of two industrial robots using a Simulink model. Then, we feed these data into our Transformer-based model and train it to be a time-series data predictor. The paper presents the experimental results that show the comparison of precision and speed of a Long-Short Time Memory (LSTM) predictor and our Transformer-based predictor.}, author = {Ma, Y. and Morozov, A. and Ding, Sheng}, journal = {The International Mechanical Engineering Congress and Exposition, IMECE}, title = {Anomaly Detection for Cyber Physical Systems using Transformers}, year = 2021 }

Forschung

Forschungsschwerpunkt: - Intelligent Anomaly Interpreter for Cyber-Physical Systems

Beschreibung: - Deep Learning-based Anomaly Detection (DLAD) methods for Cyber-Physical-Systems have shown high efficiency and accuracy over traditional anomaly detection techniques. However, the lack of interpretability has impeded
practical applications of DLAD methods. Especially for safety-critical scenarios such as Human-Robot Cooperation and Assistive Medical Robotics, it is necessary to ensure that these systems are inspected robustly and carefully. The intelligent anomaly interpreter for DLAD methods needs to solve the following problems:

• Establish trust between models and users: It is not sufficient that current DLAD models provide binary results (normal or abnormal) to safety inspectors. The reason behind a specific anomaly is also supposed to be reported to users, which can help users take further actions for the whole system. For example, feature-based anomaly interpreters can explain an anomaly via finding the most abnormal data behavior in monitored multivariate timeseries data, which can offer the reason why this point is an anomaly.

• Enable fault diagnosis for CPS: Without sufficient understanding of the decision made by DLAD models, it is almost not possible to diagnose system when an anomaly occurs. For example, for the same abnormal data behavior of multivariate time-series data generated from CPS, the root cause can be different. And vice versa, different failure modes could happen after the same fault injection action. To do fault diagnosis, expert knowledge is needed for a specific CPS.

• Reduce false positives: High rate of false positives (FP) poses a critical challenge for deploying DLAD models into practical safety-critical CPS applications. Anomaly interpreter needs to give evidence to explain whether an outlier will eventually be an anomaly or not. For example, not every wrong sensor data reported to the controller will cause a failure. Thus, the context of an anomaly also needs to be considered when we deploy DLAD methods to CPS applications.

Forschungsportal:

Google Scholar: https://scholar.google.com/citations?user=vfJJEZAAAAAJ&hl=zh-CN

ResearchGate: https://www.researchgate.net/profile/Yuliang-Ma-4