Dr. Hyun Gook Kang
Rensselaer Polytechnic Institute
Safety-critical systems engineering with software: Data analytics and software testing
CCP Room 259 or join via Zoom
Dr. Kang is a Professor of Dept. of Mechanical, Aerospace, and Nuclear Engineering at Rensselaer Polytechnic Institute (RPI). Before joining RPI, he was an Associate Professor at Korea Advanced Institute of Science and Technology (KAIST) and a senior research staff of the Korea Atomic Energy Research Institute (KAERI). He also taught at Khalifa University in UAE in 2011 and 2012.
After his PhD from Nuclear Engineering Department of KAIST in 1999, Dr. Kang’s research focus has been on innovations of dynamic risk assessment of safety-critical applications including digitalized I&C systems. The topics include digital I&C risk, passive safety features, human errors, intelligent control and protection, and advanced emergency procedures. His long-term research goal is to develop a risk-free autonomous operation scheme for nuclear power plants. He is the vice-chair of Human Factors and I&C division of American Nuclear Society and the chair of Safety review broad of RPI reactor. He authored more than 300 journal and conference articles.
In the safety-critical systems engineering, conventional computer works have been used for plant response calculations, so main target was finding numerical solutions of complex equations and boundary conditions. Enhanced computational power and sophisticated algorithms enable us to proceed more intensive utilization of advanced techniques in even the most conservative engineering field. In this talk, we will address two different aspects of computerization of safety-critical applications. One is the systematic framework for computer-based decision making and the other is the verification of embedded software algorithms.
The possible accident and mitigation scenarios produced by dynamic probabilistic risk assessment techniques, due to its dynamic and uncertain nature, would build a huge pile of data. In order to capture the risk profile and to extract other useful information from this large data set, advanced techniques of systematic data analytics must be developed. In the safety-critical applications, the data handling process must be objective and traceable so that it can serve as the common platform of data analysis for experts from various fields. Physics-informed modeling is also an important feature of this framework. With Markov-decision process, Multi-flow modeling, and AI-based data grouping, a decision-making example of nuclear power plant operation will be discussed.
On the other hand, digitalized systems can be used as a part of real-time signal processing system in safety-critical applications. The software of these systems is considered as a vulnerable point of common-cause failures which dominate the risk of whole plant, thus their testing is one of the most important issues in this field. By its nature, this kind of system software is a logical matter and determines the function of hardware in the digitalized environment. The space that digitalized input and internal variables construct can be considered as the domain that the software may encounter during system operation, which may be large but not infinite. If we can perform the software testing over the whole of this space, the limitations related to the state-of-the-art test-based software reliability quantification method, such as the uncertainty in input selection and model parameter estimation, can be resolved. In this talk, the automated exhaustive test case generation framework for NPP safety-critical software testing will be discussed.