RA-CPS

Risk Analysis of Industrial Cyber-Physical Systems

Project description

Cyber-Physical Systems (CPS) are advanced interconnected systems that are one key characteristic of the emerging trend towards Industry 4.0 in automation engineering. As with any industrial production system, risk analysis is one of the key challenges of the design of CPS. Classical risk analysis exploits several well-known methods for evaluating the dependability and resilience properties of pro-duction systems. However, Industry 4.0 implies that the Industrial CPS (sometimes also called Cyber-Physical Production Systems, or CPPS) are more complex from a structural and behavioral points of view and consist of distributed heterogeneous components. Classical methods cannot adequately describe sophisticated failure scenarios of modern highly dynamic, autonomous, and adaptive Industrial CPS. That is particularly true for Artificial Intelligence, especially Deep Learning, being employed for more broad types of safety-critical applications within systems. It is necessary then to revisit the classical Risk Assessment Methods (RAM) and extend them with the most promising advanced techniques that cover in-creased complexity.


The SI4 project is supported by the Bundesanstalt für Arbeitsschutz und Arbeitsmedizin. The main project goal is to evaluate the applicability of modern model-based RAMs for the analysis of Industrial CPS with machine learning components. Firstly, we will perform an extensive systematic review of the available methods and tools for risk assessment and evaluate the applicability of these methods for Industrial CPS. We will define the list of evaluation criteria. After that, we will assess available risk methods and tools that support these methods. As the second phase of the project, we will carry out an extensive systematic review of modern system modeling paradigms, and model-based risk analysis approaches. We will address two aspects: (i) fault-tolerance of ML components that are becoming integral parts of modern industrial CPS and (ii) application of ML methods to improve the reliability and safety of Industrial CPS. As the final phase of the SI4 project, we will define the criteria catalog for the case study. After that, we will develop the case study industrial CPS according to the defined criteria and evaluate the methods discovered in previous steps of the project.

Details

Project title: Safety-related Risk Assessment of a Cyber-Physical model
system for Industry 4.0 applications.

Project goal: Evaluation of the applicability of modern Model-based Risk
Assessment methods for the analysis of industrial Cyber-Physical Systems
with machine learning components.

Project term: 01.02.2020 + 42 months.

Link to the BAuA project page: https://www.baua.de/DE/Aufgaben/Forschung/Forschungsprojekte/f2497.html

Work packages (Arbeitspaketen)

WP1: Evaluation of the applicability of modern risk assessment methods for the industrial CPS.

Extensive systematic review of the methods and tools. Recommendations on combined risk assessment.

WP2: Evaluation of the applicability of system models for the industrial CPS and automated model-based risk analysis methods.

Extensive systematic review of the MBSE methods and tools and automated model-based risk assessment methods.

WP3: Machine Learning (ML) and CPS safety.

Extensive systematic review of (i) fault tolerance characteristics of ML components of industrial CPS and (ii) performances of ML and DL methods for error detection and mitigation.

WP4: Model-based development and risk assessment of a representative industrial CPS.

In the first part of the WP4, we will define the criteria catalog for the case study. After that, we will develop the case study industrial CPS according to the defined criteria and evaluate the methods discussed in WP1, WP2, and WP3.

Team

 

Jun.-Prof. Dr.-Ing. Andrey Morozov
Juniorprofessur, Institut für Automatisierungstechnik und Softwaresysteme

The research interest of Jun.-Prof. Morozov lies at the intersection of three domains, namely, (i) Networked Automation Systems (NAS), (ii) Dependability, and (iii) Artificial Intelligence (AI). Modern NAS is a particular case of Cyber-Physical Systems (CPS) with the focus on the cooperation of heterogeneous industrial robotic systems

E-Mail schreiben


text

 

Dr.-Ing. Silvia Vock
Project leader at BAuA (Federal Institute for Occupational Safety and Health)

Research in the field of Safety of Machinery with special focus on Risk Assessment of Safety-critical AI Applications in Industrial Automation and Cyber-Physical Systems

E-Mail schreiben


text
 
 
 
 

 

Tagir Fabarisov, M.Sc.
Institut für Automatisierungstechnik und Softwaresysteme

E-Mail schreiben

text
 
 
 
 
 

 

Georg Siedel, Dipl.-Ing.
Bundesanstalt für Arbeitsschutz und Arbeitsmedizin

E-Mail schreiben

text
 
 
 
 
 

 

Martin Westhoven, Dipl. Inform.
Bundesanstalt für Arbeitsschutz und Arbeitsmedizin

E-Mail schreiben

text

Acknowledgement

The project is supported by Bundesanstalt für Arbeitsschutz und Arbeitsmedizin.

Projektnummer: F 2497.

Zum Seitenanfang